The Effects of Immigration Restriction Laws on Immigrant

The Effects of Immigration Restriction Laws on Immigrant

Segregation in the Early Twentieth Century U.S.∗

Dafeng Xu†

Abstract

Several immigration restriction laws, passed in the background of anti-immigration populism in

the early 1920s, put an end to the period of “open borders” in the U.S. Did immigration restriction

laws affect immigrant segregation in the U.S.? In this paper, I use the linked county and individual

sample constructed based on decennial censuses in the early twentieth century to examine how

immigration restriction laws led to different trends in segregation across groups. Results shows

that all immigrant groups followed similar trends in segregation before the passage of immigra-

tion restriction laws; afterwards, segregation among immigrants from more restricted countries

declined relative to segregation among immigrants from less restricted countries. Such differ-

ences in segregation patterns were largely related to restrictions on new immigration from more

restricted countries; selection on return migration could also partially explain the results. Further

analyses suggest that de-segegation might not benefit immigrants’ assimilation.

Keywords: immigration restriction; law; immigrant segregation; early twentieth century; U.S.

∗I thank Ran Abramitzky, Martha Bailey, Leah Boustan, Giovanni Facchini, Janna Johnson, Jason Ker-win, Daniel Lichter, Matt Nelson, Matthew Notowidigdo, Sam Peltzman, Margaret Peters, Hillel Rapoport,Evan Roberts, Steven Ruggles, Matthew Sobek, David Van Riper, Hans-Joachim Voth, Zachary Ward, JohnRobert Warren, and Ariell Zimran for their comments and suggestions. I also receive helpful comments fromparticipants at numerous seminars and conferences.†225 19th Ave. S., University of Minnesota, Minneapolis, MN 55455. Email: [email protected].

1

1 Introduction

In the early twentieth century, the U.S. passed several immigration restriction laws in the

background of anti-immigration populism, putting severe limitations on immigration from

Europe, Asia, and Africa. Following the earlier Emergency Quota Act of 1921, the Im-

migration Act of 1924 marked the end of the “open borders period” in the U.S. history

(Ngai, 1999, 2004). A quota system was established and the number of newcomers could

not exceed the quota for each sending country. While quotas were assigned to all Euro-

pean countries, the effects of quota systems were heterogeneous. In particular, very low

quotas were assigned to Southern and Eastern European countries. Using five U.S. decen-

nial censuses from 1900 to 1940 (Ruggles et al., 2017), I link U.S. counties and individual

immigrants across censuses, and examine effects of immigration restriction laws on im-

migrant segregation at the county level. Specifically, I study how immigration restriction

laws led to different trends in segregation between immigrant groups that were less or not

restricted, and immigrant groups that were severely restricted.

Immigrant segregation can be dynamic over time: on one hand, immigrants spatially

assimilate to the host society (Massey and Denton, 1985); on the other hand, many new im-

migrants prefer ethnic enclave residence upon arrival (Bartel, 1989). Moreover, the spatial

and social behaviors of the native-born population could also affect immigrant segrega-

tion. In the historical context of this paper, urbanization accelerated in the U.s. since the

late nineteenth century (Ferrie, 2003), which could affect immigrant segregation because

(a) many immigrant enclaves were located in cities, and (b) immigrants who previously

lived in rural areas also started to move to cities (Hall and Ruggles, 2004). In addition, the

early twentieth century was the era of Great Black Migration, which was partially delayed

by mass immigration (Collins, 1997), but nonetheless occurred by around 1920 (Collins

and Wanamaker, 2015). African Americans affected immigrants’ social outcomes (e.g.,

Boyd and Xu, 2003; Fox and Guglielmo, 2012): in particular, internal migration of African

Americans led to racial segregation in many cities (Shertzer et al., 2016), which could

2

further change the pattern of immigrant segregation. Finally, immigrants usually faced

hostility and discrimination, especially after the passage of immigration restriction laws

(e.g., Higham, 1963; Brown and Warner, 1992; Goldin, 1994; Ngai, 1999; Haines, 2000;

Moser, 2012; Abramitzky et al., 2017), which could also affect the dynamics of immigrant

segregation. In general, immigrant segregation in the U.S. declined until 1920 but then

started to rise (Cutler et al., 2008).

Among seven major immigrant groups that had over one million people in 1920, the

British, Canadian, Irish, and Scandinavian group were less or not restricted by immigration

restriction laws,1 while the Italian, Polish, and Russian group were severely restricted. On

the other hand, other social factors (e.g., Great Black Migration) might affect all immigrant

groups. I compare county-level immigrant segregation across groups, before and after the

passage of immigration restriction laws. I find that all major immigrant groups experienced

similar trends in segregation before the 1920s. After the 1920s, however, segregation rose

among four major less restricted groups, while the degree of segregation among three major

more restricted groups became relatively lower.

Note that it is more controversial to study Polish and Russian segregation due to statis-

tical issues in pre-1930 censuses. For example, the number of people reporting Poland as

the birthplace in the 1910 census was abnormally low, making the number of Russian-born

immigrants also unreliable.2 This might be due to political situations in Eastern Europe be-

fore World War I. In this paper, the main analysis relies on the (imperfect) solution to these

statistical issues, i.e., using the mother tongue to define the Russian and Polish group when

the birthplace information are less reliable. That said, one might still worry that the results

1Although there were more than one million German-born immigrants in 1920, in this paper I do notstudy German segregation, as the effects of immigration restriction laws (and other social factors) on Germansegregation might be confounded by the effects of World War I, which should not affect segregation of othergroups that were not considered to be enemies of the U.S. in the 1910s.

2Only 34,554 people in the 1910 census reported Poland as the birthplace (the number of Russian-bornimmigrants: 1,562,045). Although this might be because Poland gained independence only in 1918, the 1900census did document 384,058 Polish-born immigrants (the number of Russian-born immigrants: 424,648).Hence, it is more likely that there were statistical errors in the 1910 census. In any of these censuses, manyPolish immigrants might report Russia as the birthplace, as they were indeed born in the Russian part of “ge-ographic Poland”. This could even affect the 1920 census, as 1920 was very close to Poland’s independence.

3

should be interpreted with caution given that Polish and Russian segregation are calculated

not as cleanly as Italian segregation. To further explore this, I redo the empirical analysis

based on the sample without Polish and Russian immigrants, and find qualitatively similar

results. This suggests that the main conclusion of this paper is not driven or affected by the

construction of the “treatment group” or data issues in U.S. censuses.

I then study the “structure” of county-level immigrant segregation by analyzing individ-

uals’ ethnic enclave residence at the sub-county level. To do so, I construct an individual-

level panel dataset by linking the 1920 immigrant census sample to the 1930 census, and

examine the effects of immigration laws on the share of immigrants of the same origin in

the enumeration district. I find that immigration restriction laws decreased ethnic enclave

residence of immigrants from more restricted countries, which is consistent with county-

level results of trends in immigrant segregation.

There are several possible mechanisms behind the effects of immigration laws on im-

migrant segregation. Researchers observe that immigrants move to less segregated areas as

a consequence of assimilation (e.g., Duncan and Lieberson, 1959; Gordon, 1964; Massey

and Denton, 1985; Bartel, 1989). Hence, immigration laws generated direct effects on

segregation, as the number of new immigrants from more restricted countries significantly

declined after the 1920s. I indeed find that differences in arrivals of new immigrants be-

tween more and less restricted groups could lead to differences in segregation patterns.

In addition, it is theoretically possible that immigration laws could indirectly affect im-

migrant segregation through spatial behaviors of old immigrants who had settled down in

the U.S. I call this as the “indirect effect”, as immigration laws did not directly restrict

old immigrants. However, natives’ attitudes towards immigrants (e.g., hostility), especially

after the passage of immigration restriction laws, could still affect old immigrants’ spatial

behaviors. To study this, I use the linked individual panel and explore two types of spatial

behaviors: internal migration and return migration. For the channel of internal migration,

the segregation pattern would change when immigrants from more restricted countries be-

4

came more (or less) likely to move within the U.S. if they previously lived in segregated

areas. For the channel of return migration, the segregation pattern would change when im-

migrants from more restricted countries became more (or less) likely to remain in the U.S.

if they previously lived in segregated areas. Again, I focus on the enumeration district (i.e.,

sub-county) level and examine whether old immigrants’ spatial behaviors were related to

local demographic characteristics. Consistent with earlier historical research (e.g., Cerase,

1974), I find that selection on return migration might partially explains trends in immigrant

segregation: immigrants from more restricted countries were more likely to stay in the

U.S. after the 1920s if they previously lived in less segregated areas, especially compared

to those from less restricted countries. On the other hand, I find no significant effects of

internal migration on immigrant segregation.

This paper adds to the literature of population economics, economic history, public pol-

icy, and urban economics by using recently available and usable full-count census data to

examine effects of immigration restriction laws on immigrant segregation, and specifically,

how the spatial behaviors of U.S. immigrants were affected. The degree of ethnic segrega-

tion rose in the early twentieth century U.S. (Logan and Parman, 2017). Black migration

(Vigdor, 2002; Collins and Wanamaker, 2014, 2015) contributed to changing segregation

patterns, but the rise in immigrant segregation also played a crucial role (Cutler et al., 2008).

In recent years, many researchers examine how immigration restriction laws affected Euro-

pean immigrants’ return migration (e.g., Greenwood and Ward, 2015; Ward, 2017). On the

other hand, less is known about the relationship between immigration restriction laws and

trends in immigrant segregation, and furthermore, the possible mechanisms behind such

trends in segregation. This paper fills the gap in the related literature.

In the remainder of this paper, Section 2 introduces the historical background. Section

3 introduces data and methods. Section 4 presents the main empirical analysis. Section 5

further discusses mechanisms behind empirical findings of this paper. Section 6 discusses

potential effects of segregation on immigrants’ assimilation. Section 7 concludes.

5

2 Background

This section introduces the historical background. I first present a brief overview of Euro-

pean immigration in the age of mass migration and immigration restriction laws in the U.S.

I then focus specifically on immigrant segregation in the early twentieth century.

2.1 European Immigration and Immigration Restriction Laws

The U.S. absorbed nearly 30 million immigrants from Europe during the age of mass mi-

gration, when the U.S. generally remained its open borders with the exception that Chinese

immigration was severely restricted by the Exclusion Act of 1882 (e.g., Kanazawa, 2005).

There were several federal laws before 1920 that affected European immigration: the Nat-

uralization Act of 1906 (and the following state-level laws) that required immigrants to

learn English (Lleras-Muney and Shertzer, 2015), the Immigration Act of 1907 that lim-

ited immigrants with disability and disease, and the Immigration Act of 1917 that imposed

literacy tests on immigrants. However, all of the above acts were more or less regulations

rather than direct restrictions on immigration, and in particular, European immigration was

not restricted based on the country of origin (van Nuys, 2002).

Researchers and the public have long discussed effects of immigration on natives’ well-

being (e.g., Borjas, 1987; Card, 1990; Borjas et al., 1992; Altonji and Card, 1994; Orrenius

and Zavodny, 2006; Ottaviano and Peri, 2012; Akay et al., 2014). Similarly, this was one of

the major issues discussed among the public in the age of mass migration: the social envi-

ronment of anti-immigration populism had been developed since the late nineteenth century

with the concern that the mass influx of uneducated and unskilled immigrants would result

in economic downturn and harm the native-born population. The Immigration Restriction

League was established in 1894, and its members believed that immigrants from Southern

and Eastern Europe were “undesirable” relative to Anglo-Saxons (Higham, 1963; Brown

and Warner, 1992; Hatton and Williamson, 1998). Despite the social environment of anti-

6

immigration populism, it was not until 1921 that the Emergency Quota Act was passed,

and it was the first law that explicitly restricted European immigration by sending country

with quotas. This act set the annual quota for any sending country to 3% of the number of

residents of that origin residing in the U.S. in 1910 based on the census. Finally, restric-

tions on Southern and Eastern European immigration became even more severe in 1924,

when a new Immigration Act was passed and the quota for any sending country was set to

2% of the number of residents of that origin residing in the U.S. in 1890. This generated

immediate and sizable effects on the number of new immigrants from restricted countries.

Table 1: Immigration Flows in the Early Twentieth Century U.S.Italy Russia Poland U.K. Canada Ireland Scandinavia

A. Population size1900 census 490,883 424,648 384,058 1,182,239 1,229,923 1,641,393 1,078,478

1910 census 1,351,055 1,562,045 34,554 1,182,209 1,254,880 1,355,740 1,264,917

1920 census 1,608,841 1,450,734 1,133,710 1,172,036 1,213,953 1,355,740 1,183,451

1930 census 1,789,588 1,197,244 1,257,449 1,229,484 1,399,034 929,429 1,119,527

1940 census 1,636,574 1,069,635 1,003,129 955,918 1,119,087 678,074 856,085B. Year of arrival, 1930Before 1900 245,088 204,647 199,602 451,099 503,813 411,353 541,859

1900 - 1909 629,055 456,244 447,188 229,694 218,318 188,827 266,401

1910 - 1919 543,363 364,145 452,566 220,711 217,308 129,798 152,529

1920 - 1924 277,372 140,348 116,534 194,473 255,300 82,812 73,377

1925 - 1929 94,065 32,344 42,209 134,234 204,229 116,724 85,384

Table 1 presents trends in U.S. immigration by country of origin in the early twentieth

century. In Panel A, I first report the size of each population in five censuses. Panel A shows

that all three “more restricted groups” that had over one million population in 1920—i.e.,

Italian, Russian, and Polish immigrants—had relatively few immigrants in 1900. However,

the number of immigrants from these groups soon became much larger, while the num-

ber of immigrants from less restricted groups remained fairly stable. Panel A also points

out crucial statistical issues concerning Russian and Polish immigrants. In particular, the

number of Polish-born immigrants is abnormally low in the 1910 census, making the num-

ber of Russian-born immigrants also less reliable. This could be because that Poland did

not gain independence before 1918, and census takers might document Polish immigrants’

birthplace as Russia even if they reported Poland as their birthplace. Since Poland’s inde-

7

pendence was close to 1920, such issues might still exist in the 1920 census.

In Panel B, I investigate the 1930 census and examine the year of arrival among im-

migrants by group. Of course, this neglects return migrants who once lived in the U.S.,

but left before 1930. However, the statistics still suggest the rough pattern of immigration

flows for each group. Panel B presents the sharp differences in immigrant cohorts between

more and less restricted groups. A large number of immigrants from more restricted coun-

tries of origin came to the U.S. during 1920 and 1924, but then immigration from these

countries became substantially lower, after the passage of immigration restriction laws. On

the other hand, trends in immigration from less restricted countries were relatively stable,

and immigration restriction laws had no clear effects on immigration from these countries.

2.2 Immigrant Segregation in the Early Twentieth Century U.S.

Immigrant segregation was determined by several major social factors in the early twentieth

century U.S. The period of 1900 to 1940 was the era of mass internal migration. Urban-

ization accelerated in the 1900s and 1910s (Ferrie, 2003; Hall and Ruggles, 2004), and

in the late 1910s the urban population as the percentage of the total population exceeded

50%. Urbanization affected immigrant segregation as many ethnic enclaves were located

in cities. These factors could potentially affect all immigrant groups.

On the other hand, immigration restriction laws generated heterogeneous effects on dif-

ferent immigrant groups through various channels. First, in general, new immigrants are

more likely to reside within ethnic enclaves (Massey and Denton, 1985; Bartel, 1989), and

thus the lack of newcomers caused by immigration laws might reduce immigrant segrega-

tion. This is the direct mechanism through which immigration laws could affect immigrant

segregation. Moreover, immigration laws might generate indirect effects on immigrant

segregation through changes in spatial behaviors of old immigrants in the U.S., even if old

immigrants were not directly targeted. Such indirect effects might be related to natives’

attitudes towards immigrants who had already in the U.S. One possible mechanism is inter-

8

Figure 1: The Largest Immigrant Group by County, 1920 and 1930

nal migration: trends in immigrant segregation could be indirectly affected by immigration

laws through internal migration, if old immigrants from countries that were targeted by

immigration laws had unique outcomes of internal migration, e.g., if they moved to less

segregated areas after immigration restriction laws were made. Another possible mecha-

nism is selective return migration. The return migration rate was high among immigrants,

and return migration was selected based on economic status (Abramitzky et al., 2014, 2017;

Ward, 2017). While trends in return migration were similar among more or less restricted

groups before immigration laws were made (Greenwood and Ward, 2015), the rate of re-

turn migration became lower among immigrants from more restricted countries afterwards

(Ward, 2017). Immigrant segregation could be affected if return migration was statistically

related to ethnic enclave residence prior to departure.

Figure 1 shows two county maps of the largest immigrant group originally from Europe

and Canada in 1920 and 1930. The maps show that immigrant groups were concentrated in

different areas in both years. While the pattern of ethnic settlements remained fairly stable,

Figure 1 still shows declines in the number of counties where Italians and Russians (both

were severely restricted) were the largest immigrant group during the 1920s.

In Appendix A, I consider a theoretical framework that analyzes three possible mecha-

nisms behind the effects of immigration restriction laws on immigrants segregation. Here,

9

Figure 2: Illustration (1): Immigration Restriction Laws and Immigrant Segregation. (a):Segregation before 1920 (Baseline); (b): Trends in Segregation, Lack of Newcomers

I briefly summarize this framework by providing three illustrative examples from Figure

2 to 4. In each figure, I first present a “baseline” pattern of segregation in the left sub-

figure: there are four sub-areas in one county, in which rectangles represent natives, circles

represent immigrant groups that were less restricted (e.g., British immigrants), and stars

represent immigrant groups that were more restricted (e.g., Italian immigrants). I then

present the specific mechanism behind one type of the effect of immigration restriction

laws on immigrant segregation in the right sub-figure.

I begin with the direct mechanism in Figure 2. Figure 2 illustrates that immigration re-

striction laws had a direct effect on, in this case, British and Italian segregation through dif-

ferences in quotas on new immigration. British segregation rose as new British immigrants

chose to reside in British enclaves, while Italian segregation remained stable, with no new

immigration from Italy after the passage of immigration restriction laws. In Figure 3 and 4

I further present two indirect mechanisms. Figure 3 shows the relative increase in British

segregation and decrease in Italian segregation, because old Italian immigrants previously

living in more segregated areas were relatively more likely to move to less segregated ar-

eas after immigration restriction laws were made. Figure 4 shows how the mechanism of

selective return migration led to the similar result: old Italian immigrants previously living

in more segregated areas were less likely to stay in the U.S. after immigration restriction

laws were made. In Appendix A, I explain that even if immigration restriction laws did not

directly target immigrants who had settled down in the U.S., such mechanisms could still

10

Figure 3: Illustration (2): Immigration Restriction Laws and Immigrant Segregation. (a):Segregation before 1920 (Baseline); (b): Trends in Segregation, Internal Migration

Figure 4: Illustration (3): Immigration Restriction Laws and Immigrant Segregation. (a):Segregation before 1920 (Baseline); (b): Trends in Segregation, Return Migration

be effective because old immigrants—in particular, old immigrants from more restricted

countries—might adjust their spatial behaviors according to natives’ attitudes towards im-

migrants (e.g., Haines, 2000).

The theoretical framework (in Appendix A) shows that immigration laws led to the de-

cline in segregation among immigrants from more restricted countries through restrictions

on new immigration. Moreover, old immigrants who had settled down in the U.S. adjusted

their spatial behaviors, and those who previously lived in less segregated areas were more

likely to stay in the country. On the other hand, without assuming the exact form of old im-

migrants’ utility function, it is theoretically impossible to predict the sign of the selection

on internal migration, and its association with trends in segregation.

To empirically study whether and how immigration restriction laws affected immigrant

segregation, I calculate the degree of immigrant-native segregation for each immigrant

group at the county level, and use enumeration districts (ED) as the subunits. An ED is

11

an area in which all households are surveyed by one census taker, which can be as large as

a major part of a county in unpopulated rural areas, or as small as a few blocks in cities. On

average, there are about 15 to 20 EDs in each county, and each ED contains approximately

1,500 residents. For a specific immigrant group in county j, I consider the dissimilarity

index Dj (Duncan and Duncan, 1955; Duncan and Lieberson, 1959; Winship, 1977):

Dj =1

2

Nj∑k=1

|nk

nj

− ekej| (1)

where Nj is the number of enumeration districts in county j, nk is the number of na-

tives in the enumeration district k, nj is the total number of natives in county j. For this

immigrant group, ek and ej can be similarly defined as nk and nj .

Table 2: County-Level Segregation by Group, 1900 - 19401900 1910 1920 1930 1940

Italian-native 0.83 0.73 0.72 0.73 0.73

Russian-native 0.85 0.69 0.74 0.70 0.70

Polish-native 0.87 0.81 0.74 0.76 0.76

British-native 0.63 0.41 0.43 0.50 0.54

Canadian-native 0.64 0.46 0.47 0.53 0.56

Irish-native 0.67 0.48 0.52 0.61 0.69

Scandinavian-native 0.65 0.53 0.53 0.56 0.57

Norwegian-native 0.73 0.62 0.63 0.67 0.69

Danish-native 0.78 0.61 0.62 0.66 0.69

Swedish-native 0.71 0.56 0.56 0.61 0.63Observations 2,819 2,960 3,065 3,099 3,106

Standard deviations are in parentheses.

Based on this index, Table 2 presents trends in segregation from 1900 to 1940. I aggre-

gate immigrants from Scandinavia, but also list each country separately. Table 2 shows that

trends were generally similar among all immigrant groups between 1900 and 1920 before

immigration laws were made: segregation first declined in the 1900s, and then remained

fairly stable during the 1910s. Although immigrants from more restricted countries had

different “starting points” of segregation in 1900, trends in immigrant segregation were

roughly parallel across all groups before 1920. This coincided with parallel trends in new

immigration before the passage of immigration restriction laws (although more restricted

12

groups did have more newcomers), as reported in Table 1. Furthermore, Greenwood and

Ward (2015) observe parallel trends in return migration between more or less restricted

groups before immigration restriction laws were made. However, trends in segregation did

change after the 1920s: there were significant differences in segregation patterns between

more and less restricted groups. In particular, the degree of segregation among immigrants

from less restricted countries increased after the 1920s, while the degree of segregation

among immigrants from more restricted countries became relatively lower.

3 Data and Methods

This section introduces data and methods. I first discuss data and present descriptive statis-

tics. I then analyze empirical strategies.

3.1 Data

I use two datasets in the main empirical analysis of this paper. First, I use a county-level

panel dataset covering from 1900 to 1940. To construct this county panel, I aggregate

demographic, social, and economic variables at the enumeration district (ED) or county

level from individual census records. The descriptive statistics of individual characteristics

by country of origin in five censuses are reported in Appendix B. I then use EDs as subunits

of county, and calculate segregation for each immigrant group. I aggregate other control

variables (e.g., the marriage rate, average occupational scores, and sex ratio) at the county

level. I cluster three Scandinavian—Norwegian, Danish, and Swedish—group into one;

these three groups had similar characteristics (see Appendix B for details). Note that all

seven major immigrant groups are involved in the county panel (and individual records

reported in Appendix B), including Polish and Russian immigrants.

Table 3 presents descriptive statistics by census year. The average age rose over time,

from 25 years old in 1900 to nearly 30 years old in 1940. The marriage rate also became

13

higher, from 35.6% in 1900 to 44.5% in 1940. The homeownership rate slightly declined,

especially after the 1920s when the Great Depression occurred. The rate of school atten-

dance at the county level remained fairly stable. Finally, there was a decline in the average

occupational score between 1910 and 1930, but then a rise between 1930 and 1940.

Table 3: County-Level Panel Data: 1900 - 19401900 1910 1920 1930 1940

Average age 24.728 25.556 26.530 27.872 29.765(3.039) (2.981) (2.968) (3.035) (2.992)

Marriage rate 0.356 0.379 0.399 0.414 0.445(0.041) (0.033) (0.034) (0.034) (0.035)

Female rate 0.475 0.475 0.482 0.483 0.487(0.044) (0.033) (0.024) (0.023) (0.020)

% homeownership 0.556 0.552 0.545 0.508 0.480(0.170) (0.154) (0.143) (0.135) (0.125)

% school attendance 0.180 0.238 0.226 0.240 0.217(0.047) (0.038) (0.034) (0.029) (0.033)

Average occupational 6.452 7.070 5.444 5.979 7.338score (1.846) (1.916) (1.280) (1.261) (1.731)


To further examine the “structure” of county-level segregation, I use a linked individual

panel dataset constructed based on the 1920 and 1930 census. I link two censuses using

the algorithm Reclink2, which is used in the LIFE-M project at the University of Michigan

(Wasi and Flaaen, 2015; Bailey et al., 2017). The linkage rate is close to 20%, which is

comparable to other linkage projects. Results of record linkage suggest that the linked sam-

ple is a representative subsample of the linkable population in the 1930 census. I discuss

the details of the linkage procedure and linkage quality in Appendix C.

Specifically, I construct this individual panel by linking between the 1920 census and

the 1930 “linkable population” who arrived in the U.S. in or before 1920.3 The advantage of

using this linked sample is that I can observe individuals across censuses and study changes

in segregation and ethnic enclave residence at the sub-county level, and then analyze the

“structure” of county-level segregation. One disadvantage of most linkage projects is that

only male respondents can be linked, as women usually change surnames after marriage.3Note that, however, I do not attempt to construct the linked sample of Russian and Polish immigrants

due to statistical issues in censuses concerning Polish and Russian immigrants (one cannot determine Polishand Russian origins in the 1920 and 1930 census).

14

That said, sub-county findings based on the individual panel could still provide additional

support for county-level results.

Table 4: Individual-Level Panel Data: 1920 - 1930All Italian British Canadian Irish Scandinavian

Age, 1920 40.018 34.820 41.498 40.180 43.862 43.877(15.143) (12.730) (15.622) (16.792) (13.759) (14.184)

Years since 22.057 15.784 23.687 20.060 24.727 25.566migration, 1920 (13.572) (8.467) (14.969) (14.735) (14.138) (13.280)

Occupational scores, 17.446 17.755 17.790 17.239 17.967 16.6741920 (13.774) (12.596) (14.775) (15.017) (13.841) (12.956)

Occupational scores, 17.964 18.635 18.600 18.043 17.134 16.6851930 (14.180) (13.057) (15.071) (15.257) (14.036) (12.991)

Married, 1920 0.687 0.673 0.708 0.683 0.658 0.713

Married, 1930 0.753 0.805 0.756 0.729 0.694 0.716

Citizenship, 1920 0.568 0.344 0.657 0.559 0.736 0.789

Citizenship, 1930 0.767 0.631 0.847 0.731 0.889 0.802

Urban status, 1920 0.726 0.836 0.751 0.703 0.871 0.529

Urban status, 1930 0.745 0.861 0.760 0.718 0.873 0.532

Farm status, 1920 0.122 0.034 0.084 0.131 0.042 0.295

Farm status, 1930 0.113 0.038 0.077 0.118 0.037 0.292

Homeownership, 1920 0.439 0.343 0.433 0.454 0.356 0.625

Homeownership, 1930 0.577 0.553 0.566 0.564 0.517 0.633

% same origin in ED, 0.095 0.161 0.036 0.082 0.056 0.0941920 (0.108) (0.146) (0.041) (0.086) (0.051) (0.090)

% same origin in ED, 0.076 0.127 0.030 0.070 0.044 0.0861930 (0.085) (0.112) (0.028) (0.072) (0.048) (0.077)

% natives in ED, 0.748 0.673 0.789 0.774 0.745 0.7661920 (0.125) (0.133) (0.112) (0.109) (0.113) (0.104)

% natives in ED, 0.783 0.722 0.815 0.800 0.763 0.7961930 (0.108) (0.146) (0.041) (0.086) (0.051) (0.098)

Segregation with native- 0.36 0.56 0.26 0.25 0.33 0.34born whites, 1920 (0.162) (0.105) (0.078) (0.091) (0.071) (0.123)

Segregation with native- 0.35 0.52 0.28 0.25 0.35 0.33born whites, 1930 (0.141) (0.093) (0.086) (0.085) (0.073) (0.117)Observations 490,349 135,346 101,099 115,078 40,080 98,746

Standard deviations are in parentheses. Standard deviations of binary variables are omitted.

Table 4 presents descriptive statistics of this linked individual panel. Italian immigrants

were relatively younger and arrived in the U.S. later, but had similar occupational scores

with immigrants from less restricted countries. Italian immigrants were less likely to be

married in 1920, but then had the highest marriage rate in 1930. In both censuses, the

Italian group had the lowest citizenship rate. Italian immigrants had somewhat similar

locational characteristics with Irish immigrants: they were more likely to live in cities, and

less likely to live in farms; they were also less likely to be homeowners. As Italians (as well

15

as Poles and Russians) were more likely to reside in cities, it is useful to study immigrant

segregation in the urban sample. I will examine this in Section 4.

In the rest of the table I study settlement patterns by country of origin. Italian immi-

grants lived in enumeration districts with significantly higher shares of immigrants of the

same origin, and lower shares of the native-born population. Italians also lived in more

segregated U.S. counties. However, Table 4 suggests significant changes in ethnic enclave

residence and segregation patterns among Italian immigrants between the 1920 and 1930

U.S. census. The Italian population experienced sharp declines in both the share of immi-

grants of the same origin at the enumeration district level, and the degree of segregation at

the county level. The magnitudes of such demographic changes in the Italian immigrant

group appeared to be largest compared to other less restricted groups.

3.2 Empirical Strategies

I first discuss the county panel dataset, in which I collapse the data to origin-county-year

cells. This is an unbalanced panel, as a small number of counties cannot be linked due to

border changes and, in particular, a few counties in latter censuses did not even exist in

earlier censuses. However, the geography of U.S. counties generally remained fairly stable

in the early twentieth century, and it is thus possible to create the panel dataset with the

highly consistent geographic units. In robustness checks I will also conduct the empirical

analysis in the panel with only consistent counties. I estimate the following specification:

Dect = α +∑

t6=1920

β(t)Se1(t > 1920) + Xctγ + τe + ηc + κt + εect (2)

where e indexes the country of origin, c indexes the county, and t indexes the census

year. Dect is the dissimilarity index of segregation between the immigrant group e and

natives, as introduced in Equation 1. Se is a binary variable indicating that the immigrant

group was the more restricted group in immigration restriction laws. Hence, Se1(t > 1920)

16

represents whether group e was the more restricted group in the specific census year t.

Xct is the vector of time-varying county characteristics. I further control for country-of-

origin fixed effects τe, county fixed effects ηc, and year fixed effects κt. Essentially, this

is a difference-in-differences framework in which I compare segregation patterns between

more and less restricted groups (represented by Se) before and after immigration restriction

laws were made (represented by 1(t > 1920)).

Equation 2 is a baseline specification. Based on this, I consider several types of spec-

ifications. In the main specification in Section 4.1, I add the following sets of controls:

(a) county-by-year fixed effects, which account for all observable and unobservable county

time-varying characteristics; (b) continent-by-year fixed effects, which account for quota

differences between Canadian immigrants (not restricted) and less restricted groups. I also

run weighted regressions base on county population characteristics.

I further consider several types of changes to sample: (a) I exclude the Russian and

Polish group (with the origin defined based on the mother tongue in some censuses) and

redo the analysis. (b) I study the sample without Irish immigrants, as Table 2 shows slightly

different trends in Irish segregation. (c) I study the sample without Canadian immigrants,

who were never restricted. (d) I exclude the 1900 census, as there were relatively fewer im-

migrants from more restricted countries in 1900. (e) I study the sample with only consistent

counties, in which county borders remained unchanged.

I finally conduct the placebo tests, in which I suppose that immigration restriction laws

had been imposed against other major immigrant groups that were less or not restricted

in actual history. I then reestimate the above specification for these “fake more restricted

groups”. If the conclusion is similar to that observed among immigrants from actual more

restricted countries, one may worry that there were other factors affecting immigrant seg-

regation, and immigration restriction laws were not the major factors, because actual more

restricted groups were not specifically targeted and affected overall.

Following the county-level analysis, I turn to analyze the structure of segregation by

17

studying ethnic enclave residence in enumeration districts using the linked individual panel.

I employ the difference-in-difference strategy by replicating Equation 2:

Diect = α + βSie1(t > 1920) + Zitγ + τi + ηct + εiejct (3)

where i indexes the individual. Sie is the indicator that i’s group was the more restricted

group in immigration restriction laws. I include individual controls and fixed effects, and

time-varying characteristics. In this specification, β reflects the effects of immigration

restriction laws on immigrants’ settlement patterns in terms of county-level segregation. I

further estimate effects of immigration restriction laws on the share of immigrants of the

same origin (%SO)iejct, where j indexes the enumeration district in which i lived:

(%SO)iejct = α + βSie1(t > 1920) + Zitγ + τi + ηct + εiejct (4)

4 Empirical Analysis: Trends in Immigrant Segregation

This section studies the main empirical question of this paper: did immigration restriction

laws affect trends in immigrant segregation in the early twentieth century U.S.? In Section

4.1, I present main findings. In Section 4.2, I conduct several robustness checks. In Sec-

tion 4.3, I use the linked individual panel sample to study effects of immigration laws on

individuals’ settlement patterns at the enumeration district (i.e., sub-county) level.

4.1 Main Results: County Panel

In Table 5, I report the main findings of this paper. I start with the full sample, in which

all seven major immigrant groups are included. Four less restricted groups were British,

Canadian, Irish, and Scandinavian immigrants, and three more restricted groups were Ital-

ian, Polish, and Russian immigrants.

18

Table 5: Immigration Restriction Laws and Immigrant Segregation, Full SampleSegregation (Dect) between Natives/Native Whites and Immigrants (of the Specific Origin)

(1) (2) (3) (4) (5) (6) (7) (8)1920 county-level avg.: 0.73 (treatment-native) 1920 county-level avg.: 0.73 (treatment-whites)

0.49 (control-native) 0.49 (control-whites)Restricted, 1900 −0.028 −0.026 −0.034 0.024 −0.029 −0.027 −0.035 0.020(pre-legislation) (0.026) (0.025) (0.032) (0.037) (0.026) (0.026) (0.032) (0.038)

Restricted, 1910 0.031 0.033 0.027 0.056 0.031 0.032 0.027 0.053(pre-legislation) (0.037) (0.037) (0.051) (0.036) (0.037) (0.037) (0.051) (0.035)

Restricted, 1930 −0.082*** −0.082*** −0.093** −0.077*** −0.083*** −0.082*** −0.093** −0.073***(post-legislation) (0.021) (0.021) (0.027) (0.014) (0.021) (0.021) (0.027) (0.015)

Restricted, 1940 −0.137*** −0.137*** −0.151*** −0.128*** −0.137*** −0.137*** −0.152*** −0.125***(post-legislation) (0.030) (0.030) (0.035) (0.022) (0.030) (0.030) (0.034) (0.022)Sample Full Consistent Excluding Weighted Full Consistent Excluding Weighted

counties Italian ct. pop. counties Italian ct. pop.Adjusted R2 0.653 0.654 0.672 0.761 0.651 0.652 0.668 0.761Observations 76,936 74,460 68,087 76,936 76,936 74,460 68,087 76,936

Standard errors are in parentheses. *: p < .05; **: p < .01; ***: p < .001.

−.2

−.1

0.1

1900 1910 1920 1930 1940(a) Immigrant−Native Segregation

−.2

−.1

0.1

1900 1910 1920 1930 1940(b) Immigrant−(Native−)White Segregation

Figure 5: Immigration Restriction Laws and Immigration Segregation: (a) Immigrants andNatives; (b) Immigrants and Native-Born Whites

From Column 1 to 4 I study segregation between immigrants and the native-born popu-

lation. I control for county-by-year fixed effects, country-of-origin fixed effects, and stan-

dard errors are clustered at the country-of-origin level. Column 1 shows that more and less

restricted groups followed similar trends in segregation before the 1920s, with no statisti-

cally significant pre-legislation coefficients. After the 1920s, however, segregation among

immigrants from more restricted countries declined relative to segregation among immi-

grants from less restricted countries. This is consistent with the descriptive findings shown

in Table 2, i.e., the passage of immigration restriction laws led to differences in segregation

patterns between immigrant groups that were more and less restricted.

19

Column 2 shows similar results when I focus only on counties with consistent borders

(and thus the county panel is now balanced). In Column 3 I exclude Italian immigrants,

and in Column 4 I run a weighted regression based on the county population. I find the

similar empirical conclusion. In general, the effect size shown in Table 5 appears to be

large relative to the average degree of county-level immigrant segregation in 1920.

The next question related to the main results is: if immigration restriction laws did lead

to the decline in segregation, who moved to immigrants’ enclaves? Prior studies show some

indirect evidence that the native-born white population—rather than African Americans—

might contribute most to trends in immigrant segregation. Cutler et al. (2008) document

low interactions between whites and African Americans in the early twentieth century.

Specifically, Collins (1997) argue that African Americans’ migration was not parallel with

European immigration. To further analyze this, I rerun the regressions from Column 5 to

8, and find very similar quantitative results. While not reported here, I find no significant

effects on immigrant-black segregation, suggesting that trends in immigrant segregation

were most associated with migration among the native-born white population.

I present illustrative results of event-study coefficients in Figure 5. In the left figure,

I focus on effects of immigration laws on immigrant-native segregation; in the right fig-

ure, I focus on segregation between immigrants and native-born whites. Figure 5 suggests

graphic evidence that (a) pre-trends are indistinguishable from zero, and (b) the degree of

segregation among immigrants from more restricted countries became relatively lower after

the passage of immigration laws.

In Section 2 and 3, I discuss the statistical issues concerning Polish and Russian im-

migrants in earlier censuses. While I do find the effects of immigration restriction laws on

immigrant segregation in Table 5, one might worry that the results should be interpreted

with caution. I redo the analysis based on the sample without Polish and Russian immi-

grants, i.e., Italian immigrants are the only more restricted group in this sample.

Figure 6 presents graphic results. Compared with the main table, Figure 6 shows dif-

20

ferent coefficients in both the pre- and post-legislation period. Indeed, Table 5 similarly

shows that the magnitude of the effect appears to be larger in the sample that excludes Ital-

ian immigrants (see Column 3 and 7). Nevertheless, Figure 6 presents similar evidence of

the effects of immigration restriction laws in the sample without Polish and Russian immi-

grants, suggesting that the main results are unlikely to be driven by data issues in censuses

or the construction of the treatment group.

I conclude the main analysis by discussing standard errors. Clustering at the country-

of-origin level is the standard strategy (Bertrand et al., 2004) in a typical difference-in-

differences setting similar to that in this paper. However, one might worry that there are

relatively few groups in the context of this paper. In particular, in the sample that excludes

Polish and Russian immigrants, there is essentially only one treatment group (i.e., Italians).

An alternative strategy is to cluster standard errors at the origin-by-year level. While not

reported here, this strategy yields similar results of pre-trends (pre-legislation coefficients

are insignificant), and actually smaller standard errors for post-legislation coefficients.

−.1

−.0

50

.05

1900 1910 1920 1930 1940(a) Immigrant−Native Segregation

−.1

−.0

50

.05

1900 1910 1920 1930 1940(b) Immigrant−(Native−)White Segregation

Figure 6: Immigration Restriction Laws and Immigration Segregation, the Italian Sample:(a) Immigrants and Natives; (b) Immigrants and Native-Born Whites

4.2 Robustness Checks

I now conduct several additional tests to check the robustness of the main results. I start

with a variety of changes to samples and specifications in Table 6. In all regressions I

21

control for county-by-year fixed effects. Results are very similar when I control for other

sets of fixed effects (e.g., including county and year fixed effects separately, plus time-

varying county controls) and cluster standard errors at different levels. Again, I first study

segregation between immigrants of the specific origin and the native-born population from

Column 1 to 4, and then study immigrant-white segregation from Column 5 to 8.

Table 6: Immigration Restriction Laws and Immigrant Segregation, Changes to SampleSegregation (Dect) between Natives/Native Whites and Immigrants (of the Specific Origin)

(1) (2) (3) (4) (5) (6) (7) (8)Immigrant-native segregation Immigrant-white segregation

Restricted, 1900 −0.021 −0.025 0.007 −0.026 −0.026 0.003(pre-legislation) (0.017) (0.023) (0.030) (0.016) (0.023) (0.030)


Restricted, 1930 −0.063** −0.082* −0.078** −0.073** −0.059* −0.082* −0.079** −0.069**(post-legislation) (0.015) (0.028) (0.025) (0.017) (0.015) (0.028) (0.026) (0.017)

Restricted, 1940 −0.102*** −0.135** −0.130** −0.121** −0.099*** −0.135** −0.131** −0.118**(post-legislation) (0.011) (0.035) (0.029) (0.020) (0.011) (0.035) (0.029) (0.020)Sample Excluding Excluding Excluding Weighted, Excluding Excluding Excluding Weighted,

Irish Canadian 1900 % citizen Irish Canadian 1900 % citizenAdjusted R2 0.758 0.632 0.677 0.759 0.759 0.631 0.675 0.759Observations 64,688 64,183 68,528 76,936 64,688 64,183 68,528 76,936


In Column 1, I exclude Irish immigrants from the sample. Descriptive statistics pre-

sented in Section 2 and 3 show that Irish immigrants might have different settlement pat-

terns in the U.S. Therefore, it is useful to conduct the robustness check in a sample without

Irish immigrants. Column 1 of Table 6 shows different pre-legislation estimates (and are

still insignificant), but post-legislation coefficients are quantitatively very similar, suggest-

ing that the inclusion of Irish immigrants should not affect the main conclusion of this

paper. In Column 2 I exclude Canadian immigrants, who were not restricted in the 1920s.

I find similar results based on this subsample. In Column 3 I drop observations in the 1900

census. The coefficient for 1910 (pre-legislation) is still insignificant, while magnitudes of

effects in the post-legislation period are similar to those in main findings. In Column 4 I run

a weighted regression, in which weights are constructed based on the citizenship rate at the

county level. I again observe quantitatively similar effects of immigration restriction laws

in this model. From Column 5 to 8, I redo the analysis using degrees of immigrant-white

22

segregation as dependent variables, and find very similar results.

Section 3 and Appendix B mention that more restricted groups are generally more likely

to reside in cities. This suggests that more and less restricted groups might have different

“starting points” of segregation in terms of urban residence. Even if two types of groups

did follow parallel trends in segregation before the 1920s, it is useful to examine the het-

erogeneous effects by urban status, and it is particularly interesting to estimate the effects

of immigration restriction laws on immigrant segregation in urban areas.

Table 7: Immigration Restriction Laws and Immigrant Segregation, Urban/Rural SampleSegregation (Dect) between Natives/Native Whites and Immigrants (of the Specific Origin)

(1) (2) (3) (4) (5) (6) (7) (8)Immigrant-native segregation Immigrant-white segregation

Restricted, 1900 −0.024 −0.113** −0.024 0.007 −0.027 −0.115** −0.026 0.002(pre-legislation) (0.031) (0.021) (0.031) (0.040) (0.043) (0.020) (0.031) (0.040)


Restricted, 1930 −0.079*** −0.065* −0.079** −0.073*** −0.079*** −0.065* −0.078*** −0.069***(post-legislation) (0.020) (0.025) (0.020) (0.013) (0.020) (0.025) (0.020) (0.013)

Restricted, 1940 −0.132*** −0.119** −0.132*** −0.122*** −0.132*** −0.120** −0.132*** −0.118***(post-legislation) (0.028) (0.030) (0.028) (0.019) (0.028) (0.030) (0.028) (0.019)Sample Urban Rural Urban Urban Urban Rural Urban Urban

consistent weighted consistent weightedAdjusted R2 0.706 0.614 0.707 0.775 0.706 0.611 0.706 0.777Observations 45,655 31,281 44,980 45,655 45,655 31,281 44,980 45,655


I begin with the urban county sample in Column 1, Table 7. I find very similar effects

of immigration restriction laws on immigrant-native segregation in urban areas. I observe

insignificant pre-legislation coefficients, and find the relative decline of segregation among

more restricted groups after the passage of immigration restriction laws. However, results

based on the rural sample are less robust (the pre-legislation coefficient in 1900 is signif-

icant). This might be because relatively fewer immigrants from more restricted countries

resided in rural areas even in 1900, and the estimates might be less reliable. I then return

to the urban sample: in Column 3 I focus on urban counties with consistent borders, and

in Column 4 I introduce weights constructed based on county populations. Both columns

show similar effects of immigration restriction laws in the urban sample. From Column 5

to 8 I repeat the exercise using immigrant-white segregation as dependent variables, and

23

find that the results remain unchanged. Hence, the main findings of this paper are at least

robust in the sample of urban counties, in which most immigrants from more restricted

countries resided in the early twentieth century U.S.

I conclude robustness checks by discussing placebo tests. In Appendix D, I conduct

several placebo tests in which I suppose that immigrants from (one of the) less restricted

countries were affected by immigration restriction laws. If one can observe similar “ef-

fects” of immigration restriction laws, then the main results reported in this section should

be unreliable because it was probably not immigration laws—but other social factors—

that led to differences in segregation patterns between more and less restricted groups after

the 1920s. Results reported in Appendix D, however, show that immigration restriction

laws had the particular effects on more restricted groups, and different trends in immigrant

segregation should be indeed associated with immigration restriction laws.

4.3 Sub-County Evidence from Individual Records

I conclude this section by presenting sub-county evidence of trends in immigrant segre-

gation based on individual panel data in the 1920 and 1930 census. The individual panel

data contain all four major less restricted groups (British, Canadian, Irish, and Scandi-

navian), and Italian immigrants who were mainly restricted. Although some studies use

individual-level panel data from 1900 to 1940 (e.g., Abramitzky et al., 2014), this idea of

data construction might not work in this paper, as the long individual panel might make

the linked sample biased towards old immigrants (who had been in the U.S. in 1900), and

thus the linked sample is no longer representative to immigrants in 1930. I also do not link

Russian and Polish immigrants due to possible inconsistency of the birthplace information.

To explore the “structure” of immigrant segregation, I examine demographic changes

by immigrant group at the sub-county level. Specifically, in Table 8, I focus on the enu-

meration district (ED) level from 1920 to 1930, and estimate Equation 3 and 4. In Column

1 and 2, I regress the share of immigrants of the same origin in the ED on the interaction

24

Table 8: Immigration Restriction Laws and Immigrant Segregation, Individuals(1) (2) (3) (4) (5) (6) (7) (8)

Ethnic enclave characteristics: Segregation measure:% same origin, ED % native, ED with native-born whites with natives

Average in 1930: 0.127 0.722 0.52 0.52Sie1(t > 1920) −0.025** −0.027** 0.025*** 0.029*** −0.049** −0.052** −0.054*** −0.057***

(0.008) (0.010) (0.004) (0.002) (0.013) (0.018) (0.013) (0.017)Controls Yes Yes Yes Yes Yes No Yes YesState FE Yes No Yes No Yes No Yes NoCounty FE No Yes No Yes No Yes No YesAdjusted R2 0.288 0.372 0.393 0.558 0.726 0.856 0.719 0.854Observations 980,698 980,698 980,698 980,698 980,698 980,698 980,698 980,698

Standard errors are clustered at the origin-by-year level and are in parentheses. *: p < .05; **: p < .01; ***: p < .001.

between the indicator of the more restrictetd group and the year dummy, with state and

county fixed effects, respectively. I find that after the 1920s, immigrants from more re-

stricted countries (in this table, Italians) who remained in the U.S. lived in EDs with lower

shares of compatriots. The effects were large relative to the average share of compatriots

(0.127). Column 3 and 4 present similar results: after the 1920s, remaining immigrants

from more restricted countries lived in EDs with higher shares of natives. From Column 5

to 8, I turn to focus on county-level immigrant segregation, and find significantly different

effects of immigration restriction laws between more and less restricted groups in terms of

county-level segregation. The sizes of the effects were large relative to the average degree

of segregation. These sub-county findings are consistent with the main conclusion.

5 Discussions: Mechanisms

In Section 4, I show that immigration restriction laws had different effects on immigrant

segregation among more and less restricted groups. There might be multiple mechanisms

behind trends in segregation. In Appendix A I present a theoretical framework that analyzes

three possible mechanisms, and illustrate them in Section 2. In this section, I empirically

test these mechanisms.

I first investigate the direct mechanism: immigration laws could directly affect trends in

immigrant segregation through restrictions on new immigration. I use the cross-sectional

1920 and 1930 census to study restrictions on new immigration. I then focus on two indi-

25

rect mechanisms related to immigrants who had been in the U.S. before 1920. Although

immigration laws did not restrict old immigrants, they might adjust their settlement patterns

after immigration laws were made. Specifically, I examine internal migration of remain-

ing immigrants, and selective return migration. To study the latter two mechanisms, I use

linked individual panel data and analyze spatial outcomes of remaining immigrants.

5.1 Restrictions on New Immigration

I first examine the direct effects of immigration laws on trends in immigrant segregation.

Immigration restriction laws had immediate and sizable impacts on the number of new

immigrants from more restricted countries. In contrast, the number of immigrants from

other countries remained fairly stable. Classical findings in labor economics and demog-

raphy point out that new immigrants are generally more likely to reside in ethnic enclaves

(Massey and Denton, 1985; Bartel, 1989; Altonji and Card, 1991), and with the increase in

years since migration, immigrants follow trajectories of “spatial assimilation” by moving

outside ethnic enclaves. If this was similarly true in the early twentieth century, then one

should expect the decline in segregation in groups where new immigration was restricted.

I study this question in Table 9. Here, I only include Italian immigrants as the restricted

group due to statistical issues regarding Russian and Polish immigrants in the 1920 census.

While not reported here, I find similar results with the inclusion of Russian and Polish

immigrants defined based on the mother tongue. On top of Table 9 I first show the average

share of immigrants of the same origin in the ED. The descriptive findings suggest Italian

immigrants resided in areas with significantly higher shares of immigrants of the same

origin, compared to less restricted groups. However, there was a sharp decline in the local

share of immigrants of the same origin between 1920 and 1930 (from 17.1% to 13.3%),

compared to that among less restricted groups.

I then regress the share of immigrants of the same origin in the ED on years since

migration (divided by 10, hence is “decades since migration”), and other individual-level

26

Table 9: Years since Migration and Ethnic Enclave Residence (1): State Controls% of Immigrants of the Same Origin, at Enumeration District (ED) Level

(1) (2) (3) (4) (5) (6)Italian British Canadian Irish Scandinavian Pooled

1920 census:Average: 0.171 0.036 0.080 0.055 0.087 0.092Decades since −0.013*** −0.003** −0.002* −0.002** −0.003*** −0.003***migration (0.003) (0.001) (0.001) (0.001) (0.001) (0.000)

(Decades since −0.008**migration)*Italy (0.002)Adjusted R2 0.181 0.228 0.418 0.224 0.267 0.305Observations 1,609,346 1,159,148 1,271,325 1,050,668 1,183,603 6,220,0901930 census:Average: 0.133 0.033 0.072 0.050 0.070 0.078Decades since −0.007*** −0.003*** −0.002*** −0.004*** −0.005*** −0.003***migration (0.001) (0.000) (0.000) (0.001) (0.001) (0.000)

(Decades since −0.005**migration)*Italy (0.001)Adjusted R2 0.169 0.189 0.428 0.204 0.206 0.278Observations 1,788,943 1,230,211 1,398,968 929,514 1,119,550 6,467,186Controls Yes Yes Yes Yes Yes YesState FE Yes Yes Yes Yes Yes Yes

Standard errors are in parentheses and are clustered at the state level. *: p < .05; **: p < .01; ***: p < .001.I report coefficients of decades since migration (i.e., years since migration are divided by 10). State FE are included.

control variables introduced in Appendix B. In Table 9, I run regressions in each group

from Column 1 to 5; I then pool the sample, and examine differences in ethnic enclave

residence between more and less restricted groups in Column 6. Results show evidence

of spatial assimilation over time: immigrants who lived in the U.S. longer also resided in

areas with lower shares of immigrants of the same origin. The rate of spatial assimilation,

reflected by the coefficient of “decades since migration”, was largest among Italian immi-

grants, i.e, the more restricted group. Column 1 also presents a sharp decrease in the rate of

spatial assimilation among Italian immigrants. This is expected, as the average years since

migration among Italian immigrants should become substantially larger after new Italian

immigration was severely restricted by immigration laws, and thus the rate of spatial as-

similation dropped. Furthermore, Column 6 shows particularly large effects of years since

migration on ethnic enclave residence among Italian immigrants, suggesting that immigra-

tion restriction laws directly affected immigrant segregation through restrictions on new

27

immigration. which is predicted by the analytical framework presented in Appendix A. In

Appendix E I redo the analysis based on county fixed effects, and find very similar results.

5.2 Internal Migration

Immigration laws in the 1920s did not put limitations on old immigrants who had settled

down in the U.S. However, old immigrants might adjust their settlement patterns after im-

migration laws came into effect. Hence, immigration laws might generate indirect effects

through changes in spatial behaviors of old immigrants. I start with the discussion of inter-

nal migration of old immigrants who remained in the U.S. during the 1920s.

Table 10: Internal Migration Rate during the 1920s, County or State LevelAll Italian British Canadian Irish Scand. Norwegian Danish Swedish Control

Migration type:Any 44.7% 52.1% 39.1% 30.0% 55.8% 41.1% 38.6% 46.1% 41.5% 41.8%

Inter-state 32.5% 38.2% 27.6% 28.1% 43.6% 30.4% 27.0% 35.3% 31.7% 30.3%

Within-state 17.9% 22.4% 15.9% 16.4% 21.4% 15.4% 15.9% 16.7% 14.4% 16.4%Observations 490,349 135,146 101,099 115,078 40,080 98,746 39,486 15,137 43,943 355,003

Table 10 presents the internal migration rate by group during the 1920s. I calculate the

internal migration rate using linked individual panel data, by comparing places of residence

in the two censuses. I consider internal migration along three dimensions: (a) any type of

migration, including both inter-state and within-state inter-county migration; (b) inter-state

migration; (c) within-state inter-county migration. Table 10 reports that the overall internal

migration rate was high among immigrants between 1920 and 1930, and was significantly

higher than that of the full U.S. population (e.g., Molloy et al., 2011). Note that, however,

the traditional measure of internal migration is based on inter-state migration, as researchers

are only able to observe whether individuals reside outside their state of birth. Indeed, only

32.5% of all immigrants moved to other states in the 1920s, which is close to the internal

migration rate documented in earlier research. Finally, for immigrants who remained in the

same state, 20% of them moved to other counties.

In Appendix A, I argue that without assuming the exact form of immigrants’ utility

28

function, it is theoretically impossible to predict the sign of the effect of immigration re-

striction laws on ethnic enclave residence (and furthermore, segregation) through internal

migration among old immigrants remaining in the U.S. It is thus an empirical question how

internal migration of remaining immigrants was associated with trends in segregation after

the passage of immigration restriction laws. To examine the effects of internal migration

on segregation, I first estimate the following specification within the more restricted group

(in this section, Italians) in the linked sample:

Dit = α + βMi1(t > 1920) + γMi + δ1(t > 1920) + Xijµ+ τi(s) + εit (5)

where Dit is the degree of segregation in i’s county of residence. Mi is a binary indi-

cator of internal migration. This is essentially a two-period difference-in-differences struc-

ture, in which I compare the degree of segregation in 1920 and 1930 between movers and

non-movers in the more restricted group. I control for individual time-varying characteris-

tics introduced in Section 3, state fixed effects, and cluster standard errors by year. I then

include immigrants from less restricted countries in the regression, and further estimate a

difference-in-difference-in-differences specification as follows:

Diet = α + βSeMi1(t > 1920) + γSeMi + δSe1(t > 1920) + λMi1(t > 1920)

+ ρSe + φMi + ψ1(t > 1920) + Xietµ+ τi(s)t + κe + εiet (6)

where e indexes the country of origin. Similar to the main empirical analysis, I cluster

standard errors at the origin-by-year level.

In Table 11 I estimate Equation 5 and 6. I first focus only on Italian immigrants. The

first panel shows that Italian immigrants who migrated within the U.S. indeed moved to

less segregated areas. However, the effect was very small. After including less restricted

groups in the second panel, I find no significant relationship between segregation and in-

ternal migration. In Appendix F, I further examine within-state migration and inter-state

29

Table 11: The Effects of Internal Migration: Overall (Any Type)Degree of segregation % same origin, ED % native, ED

(1) (2) (3) (4) (5) (6)Italian:Mi1(t > 1920) −0.007* −0.004* −0.006* −0.007** 0.004* 0.008***

(0.003) (0.001) (0.002) (0.001) (0.001) (0.000)Adjusted R2 0.092 0.367 0.108 0.169 0.184 0.326Observations 270,692 270,692 270,692 270,692 270,692 270,692All immigrants:SeMi1(t > 1920) 0.001 −0.001 −0.0001 −0.003 0.001 0.001

(0.003) (0.002) (0.004) (0.003) (0.003) (0.002)Adjusted R2 0.628 0.727 0.206 0.287 0.256 0.388Observations 980,698 980,698 980,698 980,698 980,698 980,698Controls Yes Yes Yes Yes Yes YesState FE No Yes No Yes No Yes

Standard errors are in parentheses and clustered at the origin-by-year level. *: p < .05; **: p < .01; ***: p < .001.

migration in more or less groups. Although I find significant results in a few models, re-

sults are not consistent and suggest no overall effects on internal migration on immigrant

segregation. In sum, internal migration should not be the major channel through which

immigration restriction laws affected immigrant segregation.

5.3 Return Migration

I now study return migration. The rate of return migration among European immigrants in

the early twentieth century was very high (Greenwood and Ward, 2015), especially after

the 1920s when immigrant laws were made (Ward, 2017). This could influence trends in

segregation if return migration was selective based on individual characteristics. In Ap-

pendix A, I argue that more restricted groups would become less segregated due to the

decline in ethnic enclaves, as (a) after the passage of immigration restriction laws, recent

immigrants were less likely to remain in the U.S. than immigrants who were socially and

economically more established (Ward, 2017), and (b) recent immigrants were more likely

to reside in ethnic enclaves (Bartel, 1989).

The direct way to study return migration is to obtain (a) departure records (Ward, 2017),

or (b) records in European countries that documented migratory experiences (Abramitzky

30

Table 12: The 1920 Census and the Linked Sample (Appearing in Both Censuses)Italian British Canadian Irish Scandinavian

1920 Linked 1920 Linked 1920 Linked 1920 Linked 1920 Linkedcensus sample census sample census sample census sample census sample

Segregation, 0.57 0.56 0.26 0.26 0.25 0.25 0.33 0.33 0.34 0.34with native-whites (0.106) (0.105) (0.085) (0.078) (0.097) (0.091) (0.071) (0.061) (0.123) (0.122)

Segregation, 0.57 0.56 0.27 0.27 0.26 0.26 0.33 0.33 0.35 0.35with natives (0.106) (0.104) (0.086) (0.079) (0.098) (0.091) (0.074) (0.064) (0.123) (0.123)

# of the same 333 327 63 64 141 144 84 92 120 130origin, ED level (436) (427) (122) (104) (202) (205) (99) (102) (136) (146)

# of natives, 1275 1270 1296 1294 1235 1231 1215 1224 1094 1120ED level (731) (704) (678) (664) (630) (605) (633) (619) (590) (592)

% of the same 0.165 0.161 0.035 0.036 0.080 0.082 0.052 0.056 0.088 0.094origin, ED level (0.149) (0.144) (0.039) (0.038) (0.086) (0.086) (0.052) (0.051) (0.082) (0.090)

% of natives, 0.671 0.673 0.791 0.789 0.772 0.774 0.751 0.745 0.772 0.766ED level (0.139) (0.133) (0.121) (0.112) (0.117) (0.109) (0.125) (0.113) (0.110) (0.104)

Occupational 17.275 17.755 16.991 17.740 16.397 17.239 16.399 17.966 16.007 17.472score (12.243) (12.596) (14.667) (14.677) (14.937) (15.017) (13.934) (13.841) (12.564) (12.820)Observations 955,017 135,346 606,868 99,368 595,914 115,078 460,886 40,080 663,798 43,943


et al., 2017). However, there are currently no nationwide departure records, or migratory

records for all European countries. Another indirect way to study return migration is to link

two censuses and observe the individual characteristics among those who remained in the

country (Abramitzky et al., 2014). Table 12 compares immigrants in the 1920 census and

the linked individual panel. This presents differences between all immigrants who lived in

the U.S. in 1920 and immigrants who remained in the U.S. during the 1920s, and thus the

linked individual panel excludes return migrants. The first two rows of the table show that

remaining immigrants from the more restricted countries (here, Italians) lived in slightly

less segregated counties in 1920.

To study who left the country and how return migration could affect segregation pat-

terns, one needs to examine local ethnic composition at the sub-county level, as county-

level segregation is calculated based on demographic characteristics of subunits within the

county. In the next four rows I compare ethnic composition between the 1920 census and

the linked individual panel in the ED. Results show that Italian immigrants who remained in

the U.S. lived in EDs with relatively fewer compatriots and more natives. On the contrary,

immigrants from less restricted countries who remained in the U.S. after the 1920s lived in

EDs with more compatriots. This is consistent with the theoretical prediction in Appendix

31

A, and could partially explain differences in segregation between more and less restricted

groups. Indeed, immigrants who lived in areas with fewer immigrants of the same origin

and more natives were generally more assimilated in terms of socioeconomic status, and

scholars do find evidence of negative selection of out-migration and positive selection of

in-migration (Abramitzky et al., 2014, 2017), especially after immigration laws were made

(Greenwood and Ward, 2015; Massey, 2016; Ward, 2017). I finally study differences in

occupational scores. In all immigrant groups I find that immigrants who remained in the

U.S. had higher occupational scores. This is consistent with findings that return migration

of European immigrants was negatively selected based on economic status (Greenwood

and Ward, 2015; Ward, 2017; Abramitzky et al., 2017).

I further study effects of return migration on immigrant segregation by estimating the

following specification in the 1920 census:

(%SO)iej = α + βSeRi + γSe + δRi + Xiejµ+ τi(s) + κe + εiej (7)

where i indexes the individual, e indexes i’s country of origin, and j indexes the enu-

meration district. I examine demographic variables at the sub-county level: (%SO)iej is

the share of immigrants of the same origin as i in the enumeration district j. Se is a binary

variable indicating that i’s group e was the more restricted group, and Ri is a binary vari-

able indicating that i was found in the 1930 census (i.e., i remained in the U.S. during the

1920s). Xiej is the vector of individual characteristics, and τi(s) are state fixed effects. I

cluster standard errors by country of origin. The interaction term SeRi reflects differences

between (a) all immigrants in the 1920 and immigrants who remained in the U.S. after the

1920s, and (b) immigrants from more and less restricted countries.

In Column 1, Table 13, I present the baseline estimation of Equation 7, using the share

of immigrants of the same origin in the enumeration district as the dependent variable. I

include individual controls and country-of-origin fixed effects, but do not control for any

32

Table 13: The Effects of Return Migration: the 1920 Census% same origin, ED # same origin, ED‡

(1) (2) (3) (4) (5) (6)Average, all: 0.094 171Average, Italian: 0.165 333SeRi −0.006** −0.006*** −0.006*** −7.663** −8.863** −8.791***

(0.001) (0.001) (0.001) (1.649) (2.320) (1.573)

Se 0.082*** 0.097*** 0.098*** 180.573*** 215.127*** 220.345***(0.001) (0.013) (0.015) (3.141) (21.171) (24.904)

Ri 0.003* 0.002 0.001* 2.445 −0.987 −0.801(0.001) (0.001) (0.000) (1.717) (2.012) (1.379)

Controls Yes Yes Yes Yes Yes YesState FE No Yes No No Yes NoCounty FE No No Yes No No YesAdjusted R2 0.202 0.289 0.377 0.213 0.276 0.340

Standard errors are in parentheses and are clustered at the country-of-origin level. Observations: 3,282,483.*: p < .05; **: p < .01; ***: p < .001. ‡: controlling for ED population in the regression.

geographic unit fixed effects. Results show that Italian immigrants who remained in the

U.S. lived in areas with relatively lower shares of immigrants of the same origin in 1920. I

find similar results in Column 2 in which I control for state fixed effects, and in Column 3

in which I control for county fixed effects. The magnitudes of the effects, however, appear

to be fairly small relative to the average size of ethnic enclaves. Subsequently, I investigate

the number of immigrants of the same origin from Column 4 to 6, with the additional

control of the population size of the enumeration district. I again find the similar results.

In sum, Table 12 and 13 present empirical evidence that supports the theoretical prediction

in Appendix A, i.e., selective return migration based on ethnic enclave residence did exist,

and could partially contribute to differences in immigrant segregation between more and

less restricted groups after the 1920s, although its magnitude was moderate.

6 Further Discussions

In this section, I briefly discuss potential effects of segregation on immigrants’ assimilation.

Researchers find that de-segregation is related to immigrants’ socioeconomic assimilation

(Massey and Denton, 1985), and in particular, there is a significant relationship between

33

immigrants’ de-segregation and economic outcomes when they actively make locational

choices based on individual characteristics (e.g., Bleakley and Chin, 2010). In other words,

the decline in immigrant segregation might suggest potential economic benefits. On the

other hand, immigrants might choose to stay in ethnic neighborhoods to form ethnic social

networks within small geographic areas, and further benefit from such within-group social

relationships (Munshi, 2003). Hence, de-segregation might not necessarily lead to better

outcomes for immigrants, in particular in the contexts of this paper that trends in segrega-

tion were not the consequences of immigrants’ choices, but the consequences of legislation

against immigration that could not be decided by immigrants.

The above discussions lead to a policy-relevant question: did de-segregation lead to

immigrants’ assimilation? Before immigration restriction laws were made, a major concern

among the natives—especially politicians—was that immigrants refused to assimilate into

the mainstream society, and one of the main reasons was segregation. In this section, I

investigate effects of segregation on literacy and English proficiency at the county level,

which are two major measures of immigrants’ assimilation.

Table 14: The Effects of Segregation: Literacy RateFirst-generation immigrants Second-generation immigrantsItalian Polish Russian Italian Polish Russian

(1) (2) (3) (4) (5) (6)Average, 1920: 0.808 0.892 0.898 0.988 0.989 0.993De-segregated counties, −0.032* 0.010 0.004 −0.001 0.009 0.0031900 (0.016) (0.014) (0.010) (0.007) (0.006) (0.003)

De-segregated counties, −0.003 0.010 0.007 0.001 0.001 0.0011910 (0.015) (0.019) (0.009) (0.006) (0.008) (0.003)

De-segregated counties, 0.003 0.008 0.009 0.002 0.001 0.006*1930 (0.014) (0.011) (0.009) 0.005 (0.004) (0.003)Controls Yes Yes Yes Yes Yes YesObservations 6,492 5,353 8,188 5,363 4,822 7,249

Standard errors are in parentheses and are clustered at the county level. *: p < .05; **: p < .01; ***: p < .001.

To examine this, I focus on specific immigrant groups that are restricted (i.e., Italian,

Polish, and Russian), and compare the literacy (both speak and read) rate and the ability

of English speaking between counties experiencing the rise and decline in segregation,

34

before and after the 1920s. In other words, I run difference-in-differences estimations

for counties grouped by trends in segregation. Table 14 shows results of the literacy rate

for first-generation and second-generation immigrants (literacy was not surveyed in the

1940 census): in general, I find no significant effects of segregation on the literacy rate at

the county level in both models. The only exception is that de-segregation improved the

literacy rate among second-generation Russian immigrants, but the effect size is very small.

Table 15: The Effects of Segregation: English SpeakingFirst-generation immigrants Second-generation immigrants

Italian Polish Russian Italian Polish Russian(1) (2) (3) (4) (5) (6)

Average, 1920: 0.859 0.893 0.908 0.968 0.966 0.972De-segregated counties, −0.027 0.001 0.009 0.001 −0.001 −0.0021900 (0.014) (0.013) (0.009) (0.007) (0.007) (0.005)

De-segregated counties, 0.001 0.046*** −0.023 0.004 −0.001 0.0031910 (0.017) (0.010) (0.012) (0.006) (0.009) (0.004)

De-segregated counties, 0.008 0.021* 0.016* 0.001 −0.004 −0.0011930 (0.009) (0.008) (0.006) (0.006) (0.005) (0.004)

De-segregated counties, 0.032*** 0.043*** 0.021** 0.001 −0.004 −0.0011940 (0.009) (0.008) (0.007) (0.005) (0.005) (0.004)Controls Yes Yes Yes Yes Yes YesObservations 8,151 6,901 10,280 6,610 6,018 8,855

Standard errors are in parentheses and are clustered at the county level. *: p < .05; **: p < .01; ***: p < .001.

In Table 15 I turn to examine English speaking, and find that de-segregation had posi-

tive effects on Russian immigrants’ English skills. I also observe a higher rate of English

speaking proficiency among Italian immigrants, but only in 1940. The results are less clear

among Polish immigrants (pre-trends are also significant, and have larger coefficients). No

effects are found among second-generation immigrants. In general, Table 14 and 15 sug-

gest that de-segregation might affect assimilation patterns for some immigrants, but the

overall effects of segregation on immigrants’ assimilation appear to be unclear. This is

consistent with the earlier argument: the decline in immigrant segregation was the result of

the passage of immigration restriction laws in the early twentieth century U.S., which was

not chosen by immigrants themselves; because of this, de-segregation might not necessarily

accelerate immigrants’ assimilation into the U.S. society.

35

7 Conclusion

Immigrant segregation is an important topic in population economics and urban economics.

Cutler et al., (2008) find that immigrant segregation in the U.S. declined in the first two

decades of the twentieth century, but started to rise afterwards. Was immigration legisla-

tion in the early twentieth century U.S. associated with immigration segregation? In this

paper, I focus on immigration restriction laws that came into effect in the 1920s, and study

differences in segregation patterns between more and less restricted groups.

The empirical analysis of this paper shows that both more and less restricted groups ex-

perienced similar trends in segregation before the passage of immigration restriction laws.

Afterwards, however, two types of groups followed different trends in segregation. Specif-

ically, segregation among immigrants from more restricted countries declined relative to

segregation among immigrants from less or not restricted countries. The results are robust

to changes to samples and specifications.

In addition, I discuss three possible behavioral mechanisms behind the effects of immi-

gration restriction laws on immigrant segregation. A clear explanation of the differences in

segregation patterns between more and less restricted groups after the 1920s is that immi-

gration restriction laws had different quota rules across groups, and immigration restriction

laws thus generated direct effects on immigrant segregation through restrictions on new

immigrants, who were historically more likely to reside in ethnic enclaves (Haines, 2000).

I further investigate two possible indirect mechanisms related to spatial behaviors of old

immigrants who had settled down in the U.S., through which immigration laws might indi-

rectly affect trends in segregation. First, I examine the effects of internal migration among

immigrants who had arrived in the U.S. before the 1920s. I find no significant evidence that

internal migration of remaining immigrants led to differences in segregation patterns be-

tween more and less restricted groups. Second, I find some evidence that return migration

was negatively selected in terms of ethnic enclave residence, i.e., after immigration restric-

tion laws were made, immigrants from more restricted countries were slightly more likely

36

to remain in the U.S. if they previously lived in areas with lower shares of immigrants of

the same origin (and higher shares of natives). This is similar to selective return migration

based on economic status (e.g., Ward, 2017; Abramitzky et al., 2017).

Finally, I briefly discuss consequences of the decline in segregation caused by the pas-

sage of immigration restriction laws. Focusing on the literacy rate and the rate of English

speaking proficiency, I find effects of de-segregation in some models, but the overall ef-

fects of trends in segregation on immigrants’ assimilation appear to be less clear. This

might suggest that de-segregation, as a result of the passage of immigration restriction laws

rather than immigrants’ choices, might not result in immigrants’ assimilation in the early

twentieth century U.S.

Appendix A: A Theoretical Framework of the Effects of Im-

migration Restriction Laws on Immigrant Segregation

I consider a simple analytical framework of immigrants’ settlement patterns following the

analysis in Section 2.2. Suppose that an immigrant i considers his utility Ui = ui − ui0,

where ui is i’s utility in the U.S., ui0 is the “reservation utility” for i in his home country,

and i stays in the U.S. if Ui ≥ 0. Following the theoretical model of Edin et al. (2003) and

Munshi (2003), I assume that

ui = f(ei) + zi (8)

where ei is the measure of ethnic enclave residence, say, a continuum of locational

choice such that ei ∈ (0, 1): under this setting, ei represents the share of immigrants of the

same origin at the locale. f(ei) is the utility that i gains from ethnic enclave residence, and

individual i maximizes ui subject to a “market opportunity locus” (Edin et al., 2003):

zi = µi + g(ei, yi) (9)

37

where µi reflects i’s individual characteristics unrelated to ei; g(ei, yi) measures the

utility that i gains (e.g., labor market outcomes, marriage outcomes) through interactions—

with both immigrants of the same origin and natives—in the U.S., where yi is i’s years since

migration. I assume that ∂f(ei)∂ei

> 0 and ∂g(ei,yi)∂yi

> 0. Furthermore, for g(ei, yi), I assume

that ∂g(ei,yi)∂ei|yi=0> 0, ∂g(ei,yi)

∂ei|yi=+∞< 0, and ∂g(ei,yi)

∂ei∂yi< 0.4 These assumptions are based on

that (a) immigrants have incentives to reside in ethnic enclaves (hence ∂f(ei)∂ei

is positive); (b)

immigrants socioeconomically assimilate over time (hence ∂g(ei,yi)∂yi

is positive); (c) When

immigrants are more assimilated, they are more likely to benefit from interactions with the

majority group, instead of compatriots (hence ∂g(ei,yi)∂ei∂yi

is negative). The optimal e∗i satisfies:

∂f(e∗i )

∂e∗i+∂g(e∗i , yi)

∂e∗i= 0 (10)

The above equation suggests e∗i should not exist if yi is sufficiently small, and ui is

an increasing function of ei. Therefore, newcomers are more likely to reside in ethnic

enclaves. If so, restrictions on new immigration should lead to the decrease in segregation.

Figure 2 in Section 2.2 presents illustrations of this direct mechanism.

To further study two indirect mechanisms (i.e., internal migration and return migration)

behind effects of immigration laws, I now consider a more general framework, in which

U ′i = ui−ui0−c(ei, yi). This new term c(ei, yi) represents (negative) shocks of immigration

laws against more restricted groups, where c(ei, yi) > 0. I assume that ∂c(ei,yi)∂yi

< 0 and

∂c(ei,yi)∂ei

< 0: in other words, I assume that (a) less assimilated immigrants should be

more likely to be affected; (b) immigrants living in ethnic enclaves are less exposed to

the majority group (and thus their hostility or discrimination). Now the new optimal e+i

satisfies:∂f(e+i )

∂e+i+∂g(e+i , yi)

∂e+i− ∂c(e+i , yi)

∂e+i= 0 (11)

4Upon arrival, immigrants first rely on their ethnic social networks (e.g., Edin et al., 2003; Munshi, 2003;Damm, 2009) for support, but then receive higher utility by interacting with the majority group in the hostsociety (e.g., Massey and Denton, 1985; Bartel, 1989; Bleakley and Chin, 2010); such a pattern of spatialassimilation is a function of years since immigration.

38

Is internal migration associated with ei? If yi is sufficiently small, U ′i is still an increas-

ing function of ei. This suggests recent immigrants might be better off by moving to ethnic

enclaves following the passage of immigration laws. However, if both e∗i (in Equation 10)

and e+i (in Equation 11) exist, it is theoretically impossible to compare e∗i with e+i with-

out knowing f , g, and c. In fact, one can construct different functional forms such that

e∗i > e+i or e∗i < e+i . Hence, it is unclear whether and how immigration laws could affect

segregation through the channel of internal migration.

On the other hand, it is possible to predict selective return migration. Immigrant i will

not remain in the U.S. ifU ′i < 0. Since ∂U ′i

∂yi> 0, more recent immigrants (for whom e∗i or e+i

might not exist) have lower (maximized) utility than older immigrants, and are more likely

to leave the U.S. If so, return migration should be positively related to recent immigrants

(who prefer ethnic enclave residence), and thus immigration laws should further lead to the

decrease in segregation, as ethnic enclaves decline.

Following the above analyses, Figure 3 and 4 in Section 2.2 present illustrations of two

indirect mechanisms. In the right figure of Figure 3, some immigrants from more restricted

countries (stars) moved to other areas as the response to immigration restriction laws. In

the right figure of Figure 4, immigrants from more restricted countries were more likely

to remain in the U.S. if they previously lived in less segregated areas. In both figures,

segregation among immigrants from more restricted countries became relatively lower. In

Section 5 I empirically discuss all of the above three mechanisms, and test whether they

could explain differences in segregation patterns between more and less restricted groups.

Appendix B: Individual Characteristics

In Section 3 I report descriptive statistics of the aggregate county-level data. In this ap-

pendix, I report descriptive statistics of individual census records by country of origin. In

this paper I cluster three Scandinavian populations into one; statistics of both the whole

39

Scandinavian population and each sub-population are reported.

Table 16: Descriptive Statistics of Individual Characteristics: the 1900 CensusItalian Polish Russian British Canadian Irish Scand. Norwegian Danish Swedish

Age 31.035 31.766 29.305 42.065 35.367 45.548 38.635 40.994 38.379 37.341(14.206) (14,315) (13.997) (17.095) (16.441) (16.581) (15.159) (16.046) (14.873) (14.531)

Years since 8.749 11.048 10.035 23.676 17.794 26.887 17.596 20.247 17.446 16.105migration (7.701) (8.054) (6.894) (14.922) (13.602) (15.871) (10.438) (11.686) (10.025) (9.429)

Married 0.572 0.569 0.543 0.626 0.559 0.538 0.613 0.628 0.659 0.593

Female 0.351 0.422 0.461 0.453 0.485 0.541 0.441 0.443 0.412 0.448

Homeowner 0.115 0.275 0.171 0.393 0.344 0.373 0.546 0.640 0.543 0.492

Citizen 0.628 0.720 0.784 0.865 0.800 0.903 0.832 0.821 0.832 0.838

Occupational 13.627 12.745 13.074 13.969 12.451 11.506 11.457 10.457 11.812 11.920scores (12.504) (12.207) (14.111) (14.004) (13.549) (12.695) (11.728) (11.228) (12.171) (11.863)

Urban 0.752 0.763 0.813 0.626 0.592 0.759 0.445 0.308 0.427 0.528

Farm 0.029 0.086 0.115 0.146 0.163 0.093 0.367 0.496 0.398 0.284Standard deviations are in parentheses. Standard deviations of binary variables are omitted.


Age 30.582 30.337 29.585 43.048 39.400 46.694 41.148 41.542 41.351 40.852(13.488) (13.402) (13.492) (17.610) (16.932) (16.785) (15.692) (16.755) (15.505) (15.046)


Married 0.574 0.583 0.572 0.619 0.607 0.521 0.611 0.586 0.657 0.613

Female 0.343 0.414 0.421 0.451 0.495 0.546 0.428 0.424 0.395 0.439

Homeowner 0.177 0.261 0.218 0.390 0.381 0.369 0.542 0.579 0.547 0.517

Citizen 0.574 0.552 0.671 0.852 0.834 0.905 0.837 0.821 0.838 0.847


Urban 0.772 0.811 0.865 0.706 0.687 0.830 0.523 0.423 0.478 0.596


I start with the 1900 census in Table 16. In the 1900 census, immigrants from more re-

stricted countries were significantly younger than immigrants from less restricted countries.

They also arrived in the U.S. much later. This explains the large effects of the 1924 Act

on immigration by country of origin, as the quota system used in the 1924 Act was based

on the number of immigrants by country of origin in 1890, when most immigrants from

more restricted countries had not arrived in the U.S. yet. The marriage rate was around

60% in all immigrant groups. Immigrants from more restricted countries, especially Italian

immigrants, were substantially more likely to be male, less likely to own houses, and less

40

likely to be U.S. citizens. Note that in contrast to all other groups, more than half of Irish

immigrants were female. The homeownership rate was substantially higher among Scan-

dinavian immigrants. Among all immigrant groups, Italian and British immigrants had

higher occupational scores, while Irish and Scandinavian immigrants had relatively low

occupational scores. Immigrants from more restricted countries were much more likely to

reside in cities. Immigrants from less restricted countries were more likely to live in rural

areas, and in particular, less than half of Scandinavian immigrants lived in cities. They also

had the higher rate of farm living.


Age 35.340 36.833 35.206 45.505 42.434 48.729 46.494 46.812 46.089 46.427(14.001) (13.954) (13.866) (17.245) (18.176) (15.846) (15.515) (16.259) (15.752) (14.989)


Married 0.686 0.750 0.692 0.646 0.617 0.556 0.666 0.651 0.688 0.668

Female 0.406 0.435 0.446 0.476 0.510 0.561 0.439 0.444 0.396 0.449

Homeowner 0.115 0.361 0.276 0.393 0.344 0.373 0.546 0.640 0.543 0.492

Citizen 0.628 0.736 0.522 0.865 0.800 0.903 0.832 0.821 0.832 0.838


Urban 0.846 0.829 0.884 0.757 0.723 0.863 0.564 0.471 0.534 0.626


In Table 17 I present statistics of the 1910 census. The average age of immigrants from

more restricted countries slightly declined, in contrast to that in all other groups. This sug-

gests that a large number of new (and younger) immigrants might move to the U.S. from

more restricted countries between 1900 and 1910. Similarly, I find no significant increase

in years since migration among immigrants from more restricted countries. In 1910, on

average, most of individual characteristics were similar to those in 1900. It is worth men-

tioning that Scandinavian immigrants became more likely to live in urban areas, and less

likely to lead farming lives in 1910, compared with the 1900 statistics. Many Scandinavian

immigrants were traditionally farmers after arrival, and the rise in urban residence among

Scandinavian immigrants reflects the process of urbanization.

41


Age 41.024 42.269 41.744 46.160 41.718 48.157 50.216 50.005 49.644 50.511(13.488) (12.761) (13.451) (17.610) (19.434) (16.505) (15.970) (16.579) (16.080) (15.582)


Married 0.774 0.778 0.767 0.652 0.580 0.540 0.541 0.639 0.681 0.648

Female 0.418 0.463 0.468 0.490 0.518 0.562 0.432 0.440 0.389 0.441

Homeowner 0.489 0.528 0.385 0.462 0.443 0.452 0.597 0.591 0.588 0.602

Citizen 0.515 0.513 0.631 0.666 0.553 0.707 0.752 0.739 0.777 0.751


Urban 0.878 0.863 0.902 0.803 0.767 0.890 0.631 0.558 0.595 0.685


I report descriptive statistics of the 1920 and 1930 census in Table 18 and 19, respec-

tively. It is particularly useful to compare individual characteristics in the 1920 and 1930

census, before and after immigration restriction laws came into effect. I observe the signif-

icant increase in years since migration in most immigrant groups. However, the magnitude

of the increase in years since migration was largest among immigrants from more restricted

countries. Similarly, the average age of immigrants from more restricted countries signifi-

cantly rose during the 1920s, which was opposite to the small or insignificant change in the

average age in less restricted groups. These indirectly suggest the lack of new immigrants

from Eastern and Southern European countries due to restrictions, and most immigrants

from more restricted countries who were in the U.S. after the 1920s were old immigrants

who had settled down before immigration restriction laws were passed. Immigration laws

could have long-term impacts on differences in demographic characteristics among immi-

grant groups: Table 20 exactly presents this pattern in the 1940 census.

Appendix C: Record Linkage between Two Censuses

Record linkage has recently been a widely adopted technique in social statistics and eco-

nomic history, and scholars develop various methods to link individual records across sam-

42


Age 48.178 49.916 49.513 51.679 46.648 53.197 56.363 55.641 55.583 57.020(13.481) (12.235) (12.509) (16.097) (18.423) (15.394) (14.784) (15.252) (14.951) (14.411)

Married 0.781 0.764 0.783 0.684 0.637 0.585 0.652 0.641 0.684 0.649

Female 0.426 0.474 0.474 0.502 0.536 0.570 0.437 0.444 0.395 0.449

Homeowner 0.464 0.467 0.333 0.451 0.436 0.402 0.574 0.561 0.566 0.583

Citizen 0.644 0.623 0.719 0.763 0.677 0.803 0.820 0.809 0.843 0.819


Urban 0.854 0.838 0.887 0.753 0.699 0.846 0.624 0.573 0.594 0.664

Farm 0.030 0.059 0.047 0.052 0.081 0.027 0.185 0.226 0.206 0.155Standard deviations are in parentheses. Standard deviations of binary variables are omitted.The 1940 U.S. census did not survey the year of immigration, hence I do not report years since migration in this table.

ples (e.g., Jaro, 1989; Winkler, 1990; Ferrie, 1996; Ruggles, 2008; Goeken et al., 2011;

Abramitzky et al., 2014; Collins and Wanamaker, 2015; Feigenbaum, 2017). In this pa-

per, I use the stata command Reclink2, developed by the University of Michigan (Wasi and

Flaaen, 2015; Bailey et al., 2017), to link individual immigrants in the 1920 and 1930 U.S.

census. Reclink2 is a command for multivariate linkage, which allows the comparison of

various individual characteristics between potential matches.

I first construct subsamples of “linkable populations” by immigrant group in the 1930

census (i.e., those who had been in the U.S. before 1920), and search these linkable individ-

uals in the 1920 census. I only attempt to link male immigrants between two censuses, as

women usually change their surnames after marriage. I use the following linking variables

in the linkage process: first name, last name (also with the truncated parts of the last name),

age, year of immigration, and mother tongue. In the algorithm, I set that the mother tongue

and the first letter of the last name are required to be perfectly matched between individuals

in two samples. However, a two-year window of mismatch is allowed for age and year of

immigration. The core variable of linkage is name; Reclink2 uses the bi-gram distance to

measure two strings. I include the first and last name in the algorithm separately, and also

include the truncated parts of the last name—the first and last few characters—as linking

variables. After obtaining the potential links for each 1930-census individual, I only ac-

cept the best match from the 1920 census that has a substantially higher linkscore than the

43

second-best match.5

Table 21: Checking Linkage Quality (1)Degree of segregation # same origin, ED # native, ED

Linkable Linked p-value Linkable Linked p-value Linkable Linked p-valuepopulation sample population sample population sample

Italian 0.52 0.52 1.000 228 228 1.000 1,329 1,321 <0.001(0.092) (0.091) (228) (224) (777) (777)

Observations 842,104 135,369 842,104 135,369 842,104 135,369British 0.27 0.27 1.000 55 55 0.008 1,417 1,421 0.150

(0.087) (0.076) (77) (68) (798) (784)Observations 463,796 99,368 463,796 99,368 463,796 99,368Canadian 0.25 0.25 1.000 130 129 0.078 1,354 1,357 0.235

(0.094) (0.085) (174) (168) (769) (767)Observations 449,843 115,078 449,843 115,078 449,843 115,078Irish 0.35 0.34 < 0.001 76 75 0.040 1,334 1,340 0.104

(0.081) (0.070) (92) (89) (698) (675)Observations 312,816 40,083 312,816 40,083 312,816 40,083Norwegian 0.39 0.38 < 0.001 88 83 0.008 977 943 0.003

(0.144) (0.137) (115) (96) (725) (727)Observations 163,791 39,490 163,791 39,490 163,791 39,490Danish 0.43 0.43 1.000 57 58 0.153 1,144 1,165 0.002

(0.034) (0.036) (81) (82) (14.879) (15.068)Observations 92,418 15,317 92,418 15,317 92,418 15,317Swedish 0.039 0.039 1.000 105 101 < 0.001 1,201 1,196 0.189

(0.114) (0.115) (147) (137) (747) (768)Observations 332,067 43,943 332,067 43,943 332,067 43,943


The linkage rate in each subsample is: 16.1% for Italian immigrants, 21.4% for British

immigrants, 25.6% for Canadian immigrants, 12.8% for Irish immigrants, 24.1% for Nor-

wegian immigrants, 16.6% for Danish immigrants, and 13.2% for Swedish immigrants

(overall, 16.8% for Scandinavian immigrants). The overall linkage rate is comparable to

similar linkage projects (e.g., Goeken et al., 2011; Abramitzky et al., 2014; Bailey et al.,

2017). Note that in this paper I report the “crude” linkage rate calculated based on the full

linkable population, while some papers report the linkage rate after excluding individuals

with the same names whose potential matches cannot be distinguished and selected in the

linking algorithm (e.g., Collins and Wanamaker, 2014), which will be higher than the crude

linkage rate reported here.

While it is generally impossible to link all linkable individuals, the linked sample should

be a representative subsample of the linkable population (Collins and Wanamaker, 2014;

5The highest linkscore is 1.00 (perfectly matched) and I set the lowest possible linkscore as 0.7 (the lowerthreshold). To be selected in the linked sample, the best match must have a linkscore that is 0.1 higher thanthat of the second-best match.

44

Table 22: Checking Linkage Quality (2)% same origin, ED % native, ED Occ. scores

Linkable Linked p-value Linkable Linked p-value Linkable Linked p-valuepopulation sample population sample population sample

Italian 0.127 0.127 1.000 0.725 0.722 < 0.001 18.659 18.636 0.536(0.114) (0.112) (0.119) (0.114) (12.614) (13.056)

Observations 842,104 135,369 842,104 135,369 842,104 135,369British 0.030 0.030 1.000 0.815 0.815 1.000 18.443 18.482 0.453

(0.034) (0.028) (0.110) (0.101) (14.879) (14.885)Observations 463,796 99,368 463,796 99,368 463,796 99,368Canadian 0.071 0.070 < 0.001 0.800 0.800 1.000 18.145 18.280 0.007

(0.075) (0.072) (0.101) (0.097) (15.178) (15.598)Observations 449,843 115,078 449,843 115,078 449,843 115,078Irish 0.044 0.044 1.000 0.765 0.763 0.002 16.951 17.134 0.014

(0.049) (0.048) (0.122) (0.108) (14.879) (15.068)Observations 312,816 40,083 312,816 40,083 312,816 40,083Norwegian 0.083 0.083 1.000 0.808 0.822 < 0.001 15,979 15.981 0.978

(0.070) (0.061) (0.103) (0.087) (12.803) (13.059)Observations 163,791 39,490 163,791 39,490 163,791 39,490Danish 0.047 0.048 0.035 0.830 0.830 1.000 16.556 16.587 0.791

(0.054) (0.056) (0.101) (0.097) (13.364) (13.399)Observations 92,418 15,317 92,418 15,317 92,418 15,317Swedish 0.030 0.031 < 0.001 0.816 0.817 0.008 18.443 18.598 0.003

(0.034) (0.036) (0.110) (0.105) (14.879) (15.068)Observations 332,067 43,943 332,067 43,943 332,067 43,943


Bailey et al., 2017). In other words, ideally, there should be very small differences in indi-

vidual characteristics between the linked individual panel and the full linkable population,

so that results based on the linked sample can be safely “extrapolated” to the linkable pop-

ulation in the 1930 census. In Table 21 and 22 I compare several key variables relevant to

the historical context of this paper. In Table 21 I examine the degree of segregation, the

number of immigrants of the same origin in the enumeration district, and the number of

natives in the enumeration district. In Table 22 I examine the share of immigrants of the

same origin and the share of natives in the enumeration district, and occupational scores.

In Table 21 I first compare the share of immigrants of the same origin in the enumeration

district between the linkable population in the 1930 census and the linked individual panel.

I find almost no significant differences in the degree of segregation between the linkable

population and the linked sample, with the exception of the Irish and Norwegian group.

I similarly find small differences in the number of immigrants of the same origin and the

number of natives in the enumeration district; although some differences are significant, the

magnitudes of the differences are very small. In Table 22 I focus on the share of immigrants

45

of the same origin, the share of natives in the enumeration district, and occupational scores.

Again, the differences in these characteristics between two samples are either insignificant

or of very small magnitudes. These two tables suggest the linked individual panel dataset

is a fairly representative subsample of the linkable population in the 1930 census.

Appendix D: Placebo Tests in Section 4.2

In this appendix, I present placebo tests introduced in Section 4.2. The basic idea of placebo

tests is to examine whether immigration laws indeed led to different trends in segregation

between more and less restricted groups. If so, then one should expect results of placebo

tests different from the main findings, as “fake” restrictions were “assigned” to other groups

that were less or not restricted in actual history.

−.2

−.1

0.1

.2

1900 1910 1920 1930 1940(a) British

−.1

5−

.1−

.05

0.0

5.1

1900 1910 1920 1930 1940(b) Canadian

−.1

0.1

.2

1900 1910 1920 1930 1940(c) Irish

−.2

−.1

0.1

1900 1910 1920 1930 1940(d) Scandinavian

Figure 7: Placebo Tests: “Fake” Restrictions on Other Immigrant Groups

46

Figure 7 presents the results. I run regressions similar to the model in the main table, but

assuming that British, Canadian, Irish, and Scandinavian immigrants as the specific more

restricted group. I find almost all estimates insignificant, suggesting that immigration laws

had the particular effect on the Italian immigrants. Figure 7(c) does present some significant

results for Irish immigrants. However, the shape of the trend in Irish segregation appears

to be opposite to that shown in Figure 5. At the very least, immigration restriction laws

had very different effects on segregation between Irish immigrants and immigrants from

more restricted countries in actual history. Also, In Section 4 I show that excluding Irish

immigrants from the sample does not affect the conclusion of this paper. These suggest that

after the 1920s, differences in segregation patterns between more and less restricted groups

did exist, and were very likely to be related to immigration restriction laws.

Appendix E: Additional Tests in Section 5.1

I now present an additional table following the discussion in Section 5.1, where I analyze

spatial assimilation, i.e., new immigrants started to move out and spatially assimilated with

the increase in years in migration. In Section 5.1, I control for state fixed effects in the

regressions. Table 23 presents similar results when I control for county fixed effects.

Appendix F: Additional Tests in Section 5.2

In Section 5.2, I show that internal migration—including both inter-state and within-state

migration—had no overall effects on trends in segregation by group. I now further test

inter-state migration and within-state inter-county migration separately.

In Table 24 I examine inter-state migration. In both panels, I find no significant associa-

tion between inter-state migration and the degree of segregation. I do find some significant

but small effects of migration from Column 3 to 6: Italian immigrants moved to places

with relatively higher shares of immigrants of the same origin, and lower shares of natives.

47

Table 23: Years since Migration and Ethnic Enclave Residence (2): County Controls% of Immigrants of the Same Origin, at Enumeration District (ED) Level

(1) (2) (3) (4) (5) (6)Italian British Canadian Irish Scandinavian Pooled

1920 census:Average: 0.171 0.036 0.080 0.055 0.087 0.092Decades since −0.013*** −0.002*** −0.002*** −0.001*** −0.003*** −0.002***migration (0.003) (0.000) (0.000) (0.000) (0.000) (0.001)

(Decades since −0.009***migration)*Italy (0.002)Adjusted R2 0.303 0.411 0.579 0.370 0.538 0.386Observations 1,609,346 1,159,148 1,271,325 1,050,668 1,183,603 6,220,0901930 census:Average: 0.133 0.033 0.072 0.050 0.070 0.078Decades since −0.007*** −0.002*** −0.002*** −0.002*** −0.004*** −0.002***migration (0.001) (0.000) (0.000) (0.000) (0.001) (0.000)

(Decades since −0.005***migration)*Italy (0.001)Adjusted R2 0.264 0.341 0.578 0.354 0.456 0.352Observations 1,788,943 1,230,211 1,398,968 929,514 1,119,550 6,467,186Controls Yes Yes Yes Yes Yes YesCounty FE Yes Yes Yes Yes Yes Yes

Standard errors are in parentheses and are clustered at the county level. *: p < .05; **: p < .01; ***: p < .001.I report coefficients of decades since migration (i.e., years since migration are divided by 10). County FE are included.

This actually contradicts the main findings that the degree of Italian segregation became

relatively lower over time.

I conclude the discussion of internal migration by focusing on within-state inter-county

migration in Table 25. Results show that Italian immigrants who moved within the state

generally resided in less segregated areas, as well as enumeration districts with lower shares

of immigrants of the same origin and higher shares of natives. I find less significant effects

after including less restricted groups. Note that this appears to be opposite to the effects

of inter-state migration reported in Section 5.2. In sum, although I find significant results

in some models, the overall effects of internal migration were insignificant and could not

explain the main empirical conclusion of this paper.

48

Table 24: The Effects of Internal Migration: Inter-State MigrationDegree of Segregation % same origin, ED % native, ED

(1) (2) (3) (4) (5) (6)Italian:Mi1(t > 1920) −0.002 0.002 0.001 −0.007** −0.002 0.003***

(0.001) (0.001) (0.001) (0.001) (0.003) (0.001)Adjusted R2 0.093 0.367 0.109 0.169 0.186 0.326Observations 270,692 270,692 270,692 270,692 270,692 270,692All immigrants:SeMi1(t > 1920) 0.003 0.001 0.011** 0.007* −0.006*** −0.003*



Table 25: The Effects of Internal Migration: Within-State Inter-County MigrationDegree of Segregation % same origin, ED % native, ED

(1) (2) (3) (4) (5) (6)Italian:Mi1(t > 1920) −0.013*** −0.013*** −0.018*** −0.018** 0.012** 0.013*

(0.000) (0.000) (0.001) (0.004) (0.002) (0.003)Adjusted R2 0.095 0.361 0.109 0.161 0.200 0.320Observations 167,194 167,194 167,194 167,194 167,194 167,194All immigrants:SeMi1(t > 1920) −0.003 −0.004 −0.018*** −0.018*** 0.005 0.006



References

[1] Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson. 2014. “A Nation of Immigrants: Assimi-

lation and Economic Outcomes in the Age of Mass Migration.” Journal of Political Economy, 122(3),

467 - 506.

[2] Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson. 2017. “To the New World and Back Again:

Return Migration and Upward Mobility.” Industrial and Labor Relations Review. DOI: https://

doi.org/10.1177/0019793917726981

[3] Akay, Alpaslan, Amelie Constant, and Corrado Giulietti. 2014. “The Impact of Immigration on the Well-

Being of Natives.” Journal of Economic Behavior & Organization, 103, 72 - 92.

49

https://doi.org/10.1177/0019793917726981

https://doi.org/10.1177/0019793917726981

[4] Altonji, Joseph G., and David Card. 1994. “The Effects of Immigration on the Labor Market Outcomes

of Less-skilled Natives.” in Immigration, Trade, and the Labor Market, edited by John M. Abowd and

Richard B. Freeman. Chicago: University of Chicago Press.

[5] Avitabile, Ciro, Irma Clots-Figueras, and Paolo Masella. 2013. “The Effect of Birthright Citizenship on

Parental Integration Outcomes.” Journal of Law and Economics, 56(3), 777 - 810.

[6] Bartel, Ann P. 1989. “Where Do the New U.S. Immigrants Live?” Journal of Labor Economics, 7(4), 371

- 391.

[7] Bailey, Martha J., Connor Cole, Morgan Henderson, and Catherine Massey. 2017. “How Well Do Auto-

mated Linking Methods Perform in Historical Samples? Evidence from New Ground Truth.” working

paper.

[8] Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004 “How Much Should We Trust

Differences-In-Differences Estimates?” Quarterly Journal of Economics, 119(1), 249 - 275.

[9] Bleakley, Hoyt, and Aimie Chin. 2010. “Age at Arrival, English Proficiency, and Social Assimilation

among US Immigrants.” American Economic Journal: Applied Economics, 2(1), 165 - 192.

[10] Borjas, George J. 1987. “Immigrants, Minorities, and Labor Market Competition.” Industrial and Labor

Relations Review, 40(2), 382 - 392.

[11] Borjas, George J., Richard B. Freeman, and Lawrence F. Katz. 1992. “On the Labor Market Effects of

Immigration and Trade.” in Immigration and the Workforce: Economic Consequences for the United

States and Source Areas, edited by George J. Borjas and Richard B. Freeman. Chicago: University of

Chicago Press.

[12] Boyd, Robert L., and Xiaohe Xu. 2003. “Did Retail Enterprise Among White Immigrants Benefit from

the Residential Segregation of Blacks? A Study of Large Northern Cities in the Early 20th Century.”

Social Science Quarterly, 84(4), 934 - 945.

[13] Brown, M. Craig, and and Barbara D. Warner. 1992. “Immigrants, Urban Politics, and Policing in 1900.”

American Sociological Review, 57(3), 293 - 305.

[14] Cannato, Vincent J. 2010. American Passage: The History of Ellis Island, New York: Harper Perennial.

[15] Card, David. 1990. “The Impact of the Mariel Boatlift on the Miami Labor Market.” Industrial and Labor

Relations Review, 43(2), 245 - 257.

[16] Cerase, Francesco P. 1974. “Expectations and Reality: A Case Study of Return Migration from the United

States to Southern Italy.” International Migration Review, 8(2), 245 - 262.

[17] Collins, William J. 1997. “When the Tide Turned: Immigration and the Delay of the Great Black Migra-

tion.” Journal of Economic History, 57(3), 607 - 632.

50

[18] Collins, William J., and Marianne H. Wanamaker. 2014. “Selection and Economic Gains in the Great

Migration of African Americans: New Evidence from Linked Census Data.” American Economic

Journal: Applied Economics, 6(1), 220 - 252.

[19] Collins, William J., and Marianne H. Wanamaker. 2015. “The Great Migration in Black and White: New

Evidence on the Selection and Sorting of Southern Migrants.” Journal of Economic History, 75(4),

947 - 992.

[20] Cutler, David M, Edward L. Glaeser, and Jacob L. Vigdor. 2008. “Is the Melting Pot Still Hot? Explaining

the Resurgence of Immigrant Segregation.” Review of Economics and Statistics, 90(3), 478 - 497.

[21] Damm, Anna P. 2009. “Ethnic Enclaves and Immigrant Labor Market Outcomes: Quasi-Experimental

Evidence.” Journal of Labor Economics, 27(2), 281 - 314.

[22] Dubin, Lois C. 1999. The Port Jews of Habsburg Trieste: Absolutist Politics and Enlightenment Culture.

Palo Alto: Stanford University Press.

[23] Duncan, Otis Dudley, and Beverly Duncan. 1955. “A Methodological Analysis of Segregation Indexes.”


[24] Duncan, Otis Dudley, and Stanley Lieberson. 1959. “Ethnic Segregation and Assimilation.” American

Journal of Sociology, 64(4), 364 - 374.

[25] Edin, Per-Anders, Peter Fredriksson and Olof Aslund. 2003. “Ethnic Enclaves and the Economic Success

of Immigrants: Evidence from a Natural Experiment.” Quarterly Journal of Economics, 118(1), 329 -

357.

[26] Feigenbaum, James. 2017. “Automated Census Record Linking: A Machine Learning Approach.”

manuscript.

[27] Ferrie, Joseph P. 1996. “A New Sample of Males Linked from the Public Use Microdata Sample of the

1850 U.S. Federal Census of Population to the 1860 U.S. Federal Census Manuscript Schedules.”

Historical Methods: A Journal of Quantitative and Interdisciplinary History, 29(4), 141 - 156.

[28] Ferrie, Joseph P. 2003. “Internal Migration.” in Historical Statistics of the United States, Millennial Edi-

tion, edited by Susan Carter et al. New York: Cambridge University Press.

[29] Fox, Cybelle, and Thomas A. Guglielmo. 2012. “Defining Americas Racial Boundaries: Blacks, Mexi-

cans, and European Immigrants, 1890 - 1945.” American Journal of Sociology, 118(2), 327 - 379.

[30] Goeken, Ron, Lap Huynh, Thomas Lenius, and Rebecca Vick. 2011. “New Methods of Census Record

Linking.” Historical Methods: A Journal of Quantitative and Interdisciplinary History, 44(1), 7 - 14.

[31] Goldin, Claudia. 1994. “The Political Economy of Immigration Restriction in the United States, 1890 to

1921.” in The Regulated Economy: A Historical Approach to Political Economy, edited by Claudia

Goldin and Gary D. Libecap. Chicago: University of Chicago Press.

51

[32] Gordon, Milton M. 1964. Assimilation in American Life: The Role of Race, Religion, and National Ori-

gins. New York: Oxford University Press.

[33] Greenwood, Michael J., and Zachary Ward. 2015. “Immigration Quotas, World War I, and Emigrant Flows

from the United States in the Early 20th Century.” Explorations in Economic History, 55, 76 - 96.

[34] Gross, Dominique M., and Nicolas Schmitt. 2003. “The Role of Cultural Clustering in Attracting New

Immigrants.” Journal of Regional Science, 43(2), 295 - 318.

[35] Haines, Michael R. 2000. “The Population of the United States, 1790 - 1920.” In The Cambridge Eco-

nomic History of the United States, eds., Stanley L. Engerman and Robert E. Gallman. New York:

Cambridge University Press.

[36] Hall, Patricia Kelly, and and Steven Ruggles. 2004. “Restless in the Midst of Their Prosperity: New

Evidence on the Internal Migration of Americans, 1850 - 2000.” Journal of American History, 91(3),

829 - 846.

[37] Hatton, Timothy J., and Jeffrey G. Williamson. 1998. The Age of Mass Migration: Causes and Economic

Impact, New York: Oxford University Press.

[38] Higham, John. 1963. Strangers in the Land: Patterns of American Nativism, 1860 - 1925, New Brunswick:

Rutgers University Press.

[39] Jaro, Matthew A. 1989. “Advances in Record-Linkage Methodology as Applied to Matching the 1985

Census of Tampa, Florida.” Journal of the American Statistical Association, 84(406), 414 - 420.

[40] Kanazawa, Mark. 2005. “Immigration, Exclusion, and Taxation: Anti-Chinese Legislation in Gold Rush

California.” Journal of Economic History, 65(3), 779 - 805.

[41] Lleras-Muney, Adriana, and Allison Shertzer. 2015. “Did the Americanization Movement Succeed? An

Evaluation of the Effect of English-Only and Compulsory Schooling Laws on Immigrants.” American

Economic Journal: Economic Policy, 7(3), 258 - 290.

[42] Logan, Trevon D., and John M. Parman. 2017. “The National Rise in Residential Segregation.” Journal of

Economic History, 77(1), 127 - 170.

[43] Massey, Catherine G. 2016. “Immigration Quotas and Immigrant Selection.” Explorations in Economic

History, 60, 21 - 40.

[44] Massey, Douglas S., and Nancy A. Denton. 1985. “Spatial Assimilation as a Socioeconomic Outcome.”


[45] Molloy, Raven, Christopher L. Smith, and Abigail Wozniak. 2011. “Internal Migration in the United

States.” Journal of Economic Perspectives, 25(3), 173 - 196.

[46] Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market.”

Quarterly Journal of Economics, 118(2), 549 - 599.

52

[47] Moser, Petra. 2012. “Taste-Based Discrimination Evidence from a Shift in Ethnic Preferences after WWI.”

Explorations in Economic History, 49(2), 167 - 188.

[48] Ngai, Mae M. 1999. “The Architecture of Race in American Immigration Law: A Reexamination of the

Immigration Act of 1924.” Journal of American History, 86(1), 67 92.

[49] Ngai, Mae M. 2004. Impossible Subjects: Illegal Aliens and the Making of Modern America, Princeton:

Princeton University Press.

[50] Orrenius, Pia M., and Madeline Zavodny. 2006. “Does Immigration Affect Wages? A Look at Occupation-

Level Evidence.” Labour Economics, 14(5), 757 - 773.

[51] Ottaviano, Gianmarco I. P., and Giovanni Peri. 2012. “Rethinking the Effects of Immigration on Wages.”

Journal of the European Economic Association, 10(1), 152 - 197.

[52] Ruggles, Steven. 2008. “Linking Historical Censuses: a New Approach.” History and Computing, 14(1-2),

213 - 224.

[53] Ruggles, Steven, Katie Genadek, Ronald Goeken, Josiah Grover, and Matthew Sobek. 2017. Integrated

Public Use Microdata Series: Version 7.0 [Machine-readable database]. Minneapolis: University of

Minnesota.

[54] Sammartino, Annemarie H. 2010. The Impossible Border: Germany and the East, 19141922. Ithaca:

Cornell University Press.

[55] Shertzer, Allison, Randall P. Walsh, and John R. Logan. 2016. “Segregation and Neighborhood Change

in Northern Cities: New Historical GIS Data from 1900 - 1930.” Historical Methods: A Journal of

Quantitative and Interdisciplinary History, 49(4), 187 - 197.

[56] van Nuys, Frank. 2002. Americanizing the West: Race, Immigrants, and Citizenship, 1890-1930,

Lawrence: University Press of Kansas.

[57] Vigdor, Jacob L. 2002. “The Pursuit of Opportunity: Explaining Selective Black Migration.” Journal of

Urban Economics, 51(3), 391 - 417.

[58] Ward, Zachary. 2017. “Birds of Passage: Return Migration, Self-Selection and Immigration Quotas.”

Explorations in Economic History, 64, 37 - 52.

[59] Wasi, Nada, and Aaron Flaaen. 2015. “Record Linkage using STATA: Pre-processing, Linking and Re-

viewing Utilities.” Stata Journal, 15(3), 672 - 697.

[60] Winkler, William E. 1990. “String Comparator Metrics and Enhanced Decision Rules in the Fellegi-Sunter

Model of Record Linkage.” Proceedings of the Section of Survey Research Methods, 354 - 359.

[61] Winship, Christopher. 1977. “A Revaluation of Indexes of Residential Segregation.” Social Forces, 55(4),

1058 - 1066.

53

[62] Wyman, Mark. 1993. Round-Trip to America: The Immigrants Return to Europe, 1880 - 1930, Ithaca:

Cornell University Press.

54

Documents

The Effects of Immigration Restriction Laws on Immigrant